Detection and incidence of drug-induced liver injuries in hospital: a prospective analysis from laboratory signals

Abstract

Aims

Liver damage remains the most frequent type of adverse drug reaction (ADRs) that can lead to the withdrawal of a drug from the market. The abnormal laboratory data identified by computerized hospital information systems can be used in order to improve the detection of ADRs. Our objectives were to assess the detection and incidence of drug-induced liver abnormalities in a university hospital inpatient population and to evaluate the underreporting rate of drug-induced liver injury.

Methods

We conducted a prospective study performed 1 week per month from June to October 1997. We selected patients by a computerized process using biochemistry laboratory data, based on serum enzyme values, alanine aminotransferase (over 2 fold normal) and alkaline phosphatase (over 1.5 fold normal).

Results

Among 1976 ALT and 1814 AP assays performed during the period of the study, 156 (7.9%) and 159 (8.8%) tests, respectively, fell into the selected criteria. These concerne 147 patients. Among these patients, 13 (8.8%) cases of drug-induced liver injuries were suspected. Seven cases were asymptomatic. Six cases were classified as serious by these criteria: hospitalization to investigate the cause of health status impairment (4 patients), prolongation of hospitalization (1 patient) and life-threathening (1 patient). Using the hospitalization database, the incidence of drug-induced liver injuries was estimated as 6.6 per 1000 inpatients a week. Only 1 case was reported by physicians in the same period.

Conclusions

Computerization of biochemical data would allow the development of systems to improve detection of drug-induced injury. Moreover, underreporting remains important for such potentially serious ADRs, even in a university hospital.

Introduction

Drug-induced liver injury is a potential complication of most drug therapy since the liver is central to the metabolic disposal of most drugs and xenobiotics. Despite rigorous preclinical studies and clinical trials prior to drug licensing, liver damage remains the most frequent type of adverse drug reactions (ADRs) that can lead to the withdrawal of a drug from the market [1]. The reliable design of nonhepatotoxic molecules is not yet possible [2] and the predictive value of animal and in vitro studies to identify drug hepatotoxicity before use in humans remains disappointing [3]. Recently, tolcapone and trovafloxacine were withdrawn in Europe following the recommendations of the Committee for Proprietary Medicinal Products (CPMP) because of their hepatotoxic potential [4, 5]. Most studies evaluating the incidence rate of drug-induced liver injuries focus on selective drugs or therapeutic classes [6] and on symptomatic injuries [7]. The real frequency may be somewhat higher [7]. Moreover, hepatic injury can also occur with Over The Counter (OTC) drugs such as H₂-receptor antagonists [6, 8]. In order to improve the awareness and early detection of ADRs by physicians, the use of abnormal reports retrieved from computerized hospital information systems have been tested by several authors [9, 10].

The main aim of the present study was to assess the detection and incidence of drug-induced liver abnormalities in a population of inpatients in a universitary hospital through the recording of abnormal laboratory data. We also attempted to evaluate the underreporting rate of drug-induced liver injury.

Methods

This prospective study was performed at the Rangueil University Hospital of Toulouse (South-western France) during 1 week per month from June to October 1997 (5 weeks in total). All cases of liver injuries were identified from the liver function tests performed in the central biochemistry laboratory of the hospital. A computerized process based on the serum enzyme values allowed for the identification of the patients. Thirteen departments, with 530 beds (including cardiology, endocrinology, gastroenterology, nephrology, urology, internal medicine, neurology, ophtalmology, otorhinolaryngology, pneumology, rhumatology and stomatology) from 28 medical units participated in the study. Patients hospitalized in and emergency, visceral and orthopaedic surgery departments were excluded because of confounding factors (comorbidities) as well as those hospitalized in day care services since they might be discharged before the detection of a potentially abnormal laboratory data.

Following patient selection, hospital medical records were inspected for complementary data. If necessary, the follow up of the case was carried out with the collaboration of the patient's general practitioners.

Criteria of selection of the patients

The criteria for selection of liver injury were [11]: an increase of 2N (N is the upper limit of normal range and each activity is expressed as a multiple of N) of alanine aminotransferase (ALT) serum activity, or an increase of 1.5 of alkaline phosphatase (AP) if associated with an increase of ALT or gamma glutamyl transpeptidase (GGT) or total bilirubin (TB). According to international consensus [11, 12], liver injury was designated hepatocellular when there was an increase above 2N in ALT alone or when R ≥ 5. R is the ratio of serum activity of ALT/AP. Liver injury was designated cholestatic when there was an increase above 2N in AP or when R ≥ 2. Liver injury was designated mixed when there was an increase above 2N in ALT and an increase in AP and 2 < R < 5. The criteria for exclusion were laboratory data from patients hospitalized in surgery or day care services.

Data collection

A specific form was completed for each case of abnormal liver function. This contained the following data:

• ⋆demographic characteristics: age, gender, weight,
• ⋆consumption of alcohol,
• ⋆aetiologic diagnosis of hepatic dysfunction based on clinical and biological data and results of complementary investigations (liver biopsy, virology, immunological, liver or pancreas computed tomography, liver biopsy, cardiac investigations, etc.),
• ⋆drug exposure within 30 days before admission in the study including OTC drug use (e.g. paracetamol (acetaminophen) or herbal medicine).

  All diagnosis were reviewed and confirmed by a trained gastroenterologist and two clinical pharmacologists. Demographic data are presented as mean age ± s.e. mean.

Drug causality assessment

Drug causality assessment for each case of suspected drug-induced liver injury was carried out according to international consensus of drug-induced liver injuries [12] allowing for the following criteria: time to onset and course of reaction, risk factors and screening for other causes. The score of causality was classified in 5 degrees: ‘excluded’, ‘unlikely’, ‘possible’, ‘probable’ and ‘highly probable’.

Incidence rates

For each cause of liver disturbance, the incidence of drug-related liver injuries was calculated as the ratio of the number of cases to the number of hospitalized patients during that period, as provided by the hospitalization database which contains records of every discharge. Uncertainty of association was assessed by the 95% two-sided Poisson confidence interval (CI), assuming that the occurrence of ADRs follows a Poisson distribution [13]. To determine the incidence rate of drug-induced liver injuries, we included all cases whatever their score of causality.

Cases of drug-induced liver injury

For each case, the type of injury (hepato-cellular, cholestatic or mixed) and clinical symptoms, the seriousness of ADRs according to the World Health Organization definitions [14], the labelling of ADRs on the Summary Characteristics of Product (SPC) according to French National Formulary (Vidal) [15] and the score of causality were assessed.

Underreporting rate

Underreporting was quantified by the underreporting coefficient (U) [16], calculated as the ratio between the number of events observed during the period of the study period and those spontaneously reported by the different services to our Pharmacovigilance Regional Centre.

Results

During 5 weeks of the study, 1976 ALT and 1814 AP tests were performed in the biochemistry laboratory. Overall, 156 (7.9%) and 159 (8.8%), respectively, met the inclusion criteria. These values corresponded to 147 patients. Sex ratio (male/female) was 1.5. Mean age was 59.0 ± 2.7 years.

Table 1 shows the number of patients, percentages, incidence rate as well as the confidence interval of incidence rate corresponding to several diagnosis or conditions associated to liver injuries. Among the 147 patients, 13 cases of drug-induced injuries (8.8%) were suspected. No specific cause was found in 4 patients (2.7%) and 5 patients (3.4%) were lost of follow-up. In 6 cases (4.1%), information was insufficient and in 5 cases, mixed aetiologies were suspected. According to the hospitalization data, 1964 patients were hospitalized during the period of study.

Table 1.

Data relative to 147 cases of liver injuries collected during 5 weeks in University Hospital: number of patients, percentages, incidence rate and confidence interval of incidence rate corresponding to several diagnosis or conditions associated to liver injuries.

	Number of cases	% of patients	Incidence rate (for 1000 inpatients a week)	Confidence interval of incidence rate⋆ (for 1000 inpatients a week)
Cardiac diseases	21	14.3	10.7	6.6–16.3
Alcohol intoxication	21	14.3	10.7	6.6–16.3
Obstructive	20	13.6	10.2	6.2–15.7
cholelithiasis hepatitis
Neoplastic diseases	20	13.6	10.2	6.2–15.7
Drug-induced liver	13	8.8	6.6	3.5–11.3
disturbances
Bacterial infections	13	8.8	6.6	3.5–11.3
Viral infections	6	4.1	3.1	1.1–6.6
Auto-immune hepatitis	3	2.0	1.5	0.3–4.4
Parenteral alimentation	3	2.0	1.5	0.3–4.4
Paget's disease	2	1.4	1.0	0.1–3.6
Crohn's disease	1	0.7	0.5	0–1.8
Hepatic amylosis	1	0.7	0.5	0–1.8
Hypovolaemic shock	1	0.7	0.5	0–1.8
Regenerative nodulus	1	0.7	0.5	0–1.8
hepatitis
Insulin-dependent-	1	0.7	0.5	0–1.8
diabetes
No specific causes	4	2.7	Not applicable	Not applicable
Mixed aetiologies	5	3.4	Not applicable	Not applicable
Lost of follow up	5	3.4	Not applicable	Not applicable
Insufficient information	6	4.1	Not applicable	Not applicable

^⋆Two sided Poisson confidence intervals (CI).

Drug-induced liver injuries

Table 2 summarizes the data concerning suspected cases of drug induced-liver injury. Seven cases were asymptomatic. About half (6/13) were defined as serious according to the criteria: hospitalization to investigate the cause of health status impairment (4 patients), prolongation of hospitalization (1 patient) and life-threathening (1 patient). In 8 cases, one or several medicines taken by the patients, corresponding to 14 different active compounds, were labelled in the SPC of the French National Formulary (Vidal).

Table 2.

Data relative to the cases of drug-induced liver injury including suspected drugs, types of liver injury, clinical symptoms, criteria of serioussness, labellings in French SPC and causality scores.

Drugs	Liver injury	Clinical symptoms	Seriousness	Labelling SPC	Causality score
Amineptine	Cholestatic	Abdominal pains + hepato-cellular failure	Life threathening	Labelled	Possible
Chlorpromazine + Haloperidol + clomipramine	Mixed	Abdominal pains	Hospitalization	Labelled	Possible
Clonazepam	Mixed	No	Prolongation of hospitalization	Unlabelled	Possible
Fenofibrate	Cholestatic	No	Non serious	Labelled	Possible
Fluindione	Mixed	Vomiting + malaise + hepato-cellular failure	Hospitalization	Labelled	Possible
Ketoprofen	Mixed	No	Non serious	Unlabelled	Possible
Sulphasalazine	Hepato-	Nausea + vomiting + diarrhoea + abdominal pains	Hospitalization	Unlabelled	Possible
Sulphasalazine	Mixed	No	Non serious	Unlabelled	Possible
Alimemazine + levomepromazine + Trimipramine	Cholestatic	No	Non serious	Unlabelled only for alimemazine	Possible
Fluoxetine + clorazepate	Mixed	No	Non serious	Labelled	Possible
Diethylstilbestrol + cyproterone acetate	Cholestatic	No	Non serious	Labelled	Unlikely
Isoniazide + Pyrazinamide + rifampicin	Cholestatic	No	Non serious	Labelled	Unlikely
Quinapril	Cholestatic	Jaundice	Hospitalisation	Unlabelled	Unlikely

According to these data, the incidence of drug-induced liver injuries in hospital was 6.6 per 1000 inpatient-weeks in 1997. This means that 160 cases of drug-induced liver injuries were expected in 1997. In fact, 50 cases were reported to the Regional Pharmacovigilance Centre from 1 January to 31 December.

Underreporting rate

Among the cases of drug-induced liver injuries identified in our survey between June and December 1997, only one, a cholestatic injury with amineptine was spontaneously reported by the physicians of the hospital to the Regional Pharmacovigilance Centre. The coefficient of underreporting (U) was therefore estimated to be 13.

Discussion

Laboratory data analysis can be a tool for improving the recognition of ADRs, particularly those that remain asymptomatic. Azaz et al.[17] have developed a detection system for potential ADRs based on signals generated from the laboratory data of hospitalized patients. The use of computerized laboratory data has been advocated by several teams [18–20] who could identify 29–89% of ADRs depending on the sophistication of the system [10, 21]. Recently, Levy et al.[22] have shown that a relatively simple and easy laboratory system identifies about 60% of ADRs. However, these studies are not organ-specific and include haematological, metabolic or electrolytic disturbances, hepatic or nephrologic injuries and toxic drug level measurements. Except an abstract published by Salame et al.[23], no study focusing on the recognition and determination of the incidence of hepatic ADRs from laboratory data has been performed. The frequency can be estimated from the data cited previously. In a prospective observational study of 192 hospitalized patients, Levy et al.[17] identified 112 disturbed liver function tests. Four (occurring in 2 patients) were related to drugs (1%). None was recognized by physicians. In another retrospective study on 294 hospitalized patients, 28 cases of liver injuries were related to drug intake (9.5%) [24]. In our survey, we detected 13 cases of drug-liver injury among 1964 hospitalized patients (0.6%). The difference in frequency might be explained by several factors such as the design of the study (retrospective or prospective), the specificity of the survey on drug-induced liver injury, exhaustive recording of abnormal liver analysis and the method used for causality assessment. In our study, we defined cases according to the criteria of selection of liver injury and excluded all cases of isolated elevation of liver enzymes. Furthermore, this survey was undertaken in summer and early automn and thus might not reflect the overall potential for drug-induced hepatic disorders. The seasonal prescription of some medicines (e.g. antihistaminics, antibiotics) could explain the discrepancies in comparison with other studies.

Our data show a high underreporting rate for potentially serious ADRs, even in a university hospital. Similar results were reported by Levy et al.[22]. Underreporting is a well-known phenomenon in spontaneous reporting system, its rate depending on the characteristics of ADRs, particularly its seriousness and its labelling on the SPC [25, 26]. Our results show that half of ADRs were serious and hepatic disorders were unlabelled for 5 compounds (clonazepam, ketoprofen, sulphasalazine and quinapril) in the French National Formulary [15]. However, the potential for hepatic ADRs with ketoprofen or quinapril is listed in the Physician's Desk Reference [27]. In British National Formulary, such ADRs are described as a general comment of pharmacological classes (e.g. non steroidal anti-inflammatory drugs and angiotensin-converting enzyme inhibitors) [28]. The absence of clinical symptoms and the labelling on the SPC for most of the suspected drugs could explain the high rate of underreporting despite the seriousness of ADRs in half the cases. Since 1995, the reporting of serious and unlabelled ADRs to the Regional Pharmacovigilance Centre is compulsory in France [29]. Despite the nonreporting of ADRs, the hypothesis of drug-induced liver injury was evoked in the majority of cases (except case 11) and led to withdrawal of the suspected drug(s) by the physicians and to the regression of ADRs.

This study has allowed us to determine the incidence of drug-induced liver injury in hospitalized patients through data analysis from laboratory recording. Despite the exhaustive laboratory recording and the good quality of clinical data found in medical records, it seems difficult to extend this approach to routine practice. The development of computerization of biochemical data could improve the detection of drug-induced injury. The same procedure could be applied in other laboratories such as haematological laboratories in order to improve the recognition and reporting of serious ADRs, such as agranulocytosis or aplasia. Since the wide use of H₂-receptor antagonists or herbal remedies as OTC drugs is well-known to potentially induce asymptomatic liver disturbances or hepatitis [6, 7, 30–32], a similar study in general practice would be of interest.