Drug-induced liver injury in a Swedish University hospital out-patient hepatology clinic


Dr E. Björnsson, Department of Internal Medicine, Sahlgrenska University Hospital, SE-413 45 Gothenburg, Sweden.
E-mail: einar.bjornsson@medic.gu.se




Limited data exist on the proportion of drug-induced liver injury among out-patients seen in a hepatology clinic.


To determine the proportion of drug-induced liver injury cases, and identify the most important agents and the nature of the liver injury.


A computerized diagnoses database in an out-patient hepatology clinic in a Swedish University hospital was analysed during the period 1995–2005. All suspected drug-induced liver injury cases were causality assessed with the International Consensus Criteria.


A total of 1164 cases were seen for the first time during this period. Drug-induced liver injury with at least a possible causal relationship was found in 77 cases (6.6%), 38 (3.3%) of whom were referred for evaluation to the out-patient clinic whereas 3% had a follow-up after hospitalization of drug-induced liver injury. The median age was 58 years, 43 (56%) were females, a hepatocellular pattern was observed in 37 cases (48%), cholestatic in 31 (40%) and mixed in 12%. Antibiotics were the most common agents causing drug-induced liver injury followed by non-steroidal anti-inflammatory drugs, with diclofenac most often responsible for the drug-induced liver injury.


Drug-induced liver injury cases constituted 6% of all out-patients and 3% of referrals and occurred more often in women. Antibiotics and diclofenac were the most common causes of drug-induced liver injury among out-patients.


Hepatotoxicity because of drugs is an uncommon entity in the general population.1–3 Drug-induced liver injury (DILI) caused by idiosyncratic drug reactions is in most cases detected in a patient with mild to moderate elevations of liver tests, but can in rare cases lead to fatal hepatic necrosis.4–6 Studies before the release of a drug are usually underpowered to detect rare side effects such as DILI and the information on these is usually obtained in the postmarketing phase. In fact, hepatotoxicity is the main reason for postmarketing regulatory decisions, including drug withdrawal.7 However, the real incidence of DILI remains unknown because of the difficulty in establishing diagnosis and the low reporting frequency to the pharmacovigilance authorities. The difficulty to establish a correct diagnosis was emphasized in a retrospective study from England which found that more than half of reports were unrelated to drug hepatotoxicity.8 Most of the studies on DILI are based on the spontaneous reporting to the regulatory agencies9–11 or from case reports. A careful prospective search for DILI among the French population has been reported which revealed a low spontaneous reporting of DILI.12 Recently, the frequency of DILI was reported from a tertiary referral centre over a 10-year period13 and represented 0.8% of all out-patient hepatology referrals. In Sweden, hepatic injury due to drugs occurred in 2.3% of patients hospitalized for severe jaundice14 but the incidence of DILI in the out-patient clinic is unexplored. The aims of the present study were to determine the incidence of DILI in the out-patient hepatology clinic and which drugs are responsible as well to determine the proportion of cases reported to the Swedish Adverse Drug Reaction Advisory Committee (SADRAC).

Material and methods

The Sahlgrenska University Hospital receives patients from an urban area of Gothenburg with approximately 300 000 inhabitants. All patients with a diagnosis of liver disease admitted to the out-patient hepatology were identified from a computerized diagnoses database at our institution covering the period 1995–2005. The medical records from all these patients were retrospectively reviewed. Apart from the cases identified from the diagnoses database, DILI cases during this period were also looked for in the voluntary-based reporting of adverse liver reactions that exist in our hospital and only cases reported from our unit were included in the analysis. Cases associated with acute liver injury with coagulopathy (International Normalized Ratio (INR): >1.5) or/and those who led to liver transplantation were seen at the out-patient clinic after the transplantation were excluded. Patients with other liver diseases were included in the survey with the aim to describe the real clinical setting.

For each case with suspected DILI the following data were collected: age, gender, primary referral or follow-up after hospitalization, duration of treatment, type of DILI [hepatocellular (HC), cholestatic (CS), mixed] and time to biochemical resolution. The biochemical data collected were: bilirubin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP) and PK levels.

Causality assessment was based on the International Consensus Criteria.15, 16 Thus, causal relationship between the drug and the liver reaction was classified as highly probable, probable, possible, unlikely or excluded.


The Fisher exact test was used to test differences between groups regarding dichotomous variables; the Mann–Whitney test was used for continuous variables. All test were two-tailed and were conducted at 5% significant level.


During the study period a total of 1164 cases were seen as out-patients at the hepatology out-patient clinic. Alcoholic liver disease (n = 223, 19%), non-alcoholic fatty liver disease (n = 196, 17%) and hepatitis C (n = 147, 12.6%) were the most common diagnoses. The other diagnoses encountered were PBC (9%), AIH (9.2%), cryptogenic cirrhosis (6%), undetermined cause (8%) and other diagnoses (12.4%). Most of the patients with alcoholic liver disease had a follow-up after hospitalization due to complications of liver cirrhosis and were not referred for evaluation of abnormal liver tests (data not shown). Among the cases identified through the voluntary reporting system in our hospital, the vast majority had been reported from the out-patient clinic of the Section of Gastroenterology and Hepatology and were identified by the search of diagnoses database. Only a very few cases of DILI were reported from other departments in the hospital and these cases were excluded from the final analysis. However, among the patients who were referred for investigation of liver tests abnormalities most patients had NAFLD (data not shown). A total of 114 patients had been classified as having DILI leading to a withdrawal of a number of drugs. After a careful review of their medical records, 34 cases were excluded. In a total of 23 cases the diagnosis was changed after further investigation to autoimmune hepatitis (AIH),7 non-alcoholic fatty liver disease,5 PSC,3 PBC,3 alcoholic hepatitis,2 choledocholithiasis1 and liver metastasis from carcinoid tumour.1 In another 10 cases a causal relationship with drugs was classified as unlikely or excluded according to the International Consensus Criteria and in two cases the medical records were incomplete.

Drug-induced liver injury with at least a possible causal relationship was found in 77 cases (6.6%) among the 1164 out-patients cases. The total number of 77 valid cases, gave a crude incidence of 2.3 cases per 100 000 inhabitants per year during the study period. The distribution of patients with suspected DILI during the study period is illustrated in Figure 1. A total of 38 (3.3%) of these were referred for evaluation to the out-patient clinic for abnormal liver tests, whereas 39 (3.3%) had a follow-up after hospitalization of DILI. A highly probable causal relationship was found in 26 cases (34%), probable in 35 (45%) and possible in 16 cases (21%). Only 44 (57%) of these cases were reported to the SADRAC.

Figure 1.

 The distribution of drug-induced liver injury cases during the study period.

Among the DILI patients the median age was 58 (IQR: 41–68; range: 16–86) and 43 (56%) were of female gender. A total of 49 (64%) patients were older than 50 years of age. Among patients older than 50 years, a similar proportion of females and males were observed (65% vs. 62%; P = N.S.).

A HC pattern was observed in 37 cases (48%), CS in 31 (40%), mixed in nine patients (12%) and 29 patients (38%) had jaundice [bilirubin ≥2 upper limit of normal (ULN)]. The median age among patients with HC pattern was 53 years (38–66) vs. 60 years (48–70) in patients with CS/mixed pattern (P = N.S.). Females had a higher frequency of HC type of DILI in comparison with males [27 of 43 (63%) vs. 10 of 34 males (29%)] and a lower frequency of CS/mixed pattern (P = 0.007; Figure 2). This was not attributable to differences in consumption of drugs considered responsible to the reaction. For example, there was an even sex distribution among the cases attributed to flucloxacillin and diclofenac, the drugs most often associated with a CS or mixed pattern of reaction.

Figure 2.

 The proportion of hepatocellular type of drug-induced liver injury as well as cholestatic and mixed in males and females.

Antibiotics were the most common agents associated with DILI with a total of 23 of 77 cases (30%). The antibiotics most often encountered were flucloxacillin (eight cases), ciprofloxacin (four cases), macrolides (erythromycin, azitromycin and clarithromycin), nitrofurantoin (two cases), clindamycin (two cases) as well as amoxicillin/clavulanic acid, doxycyclin, loracarbef and moxifloxacin, with one case each, respectively. DILI was because of non-steroidal anti-inflammatory drugs (NSAIDs) in 17 cases (22%, nine females) with diclofenac (14 cases) most often responsible for the DILI and one case each with rofecoxib, naproxen and ibuprufen. The drugs with at least two cases associated with DILI are illustrated in Figure 3 and the clinical characteristics of drugs associated are presented in Table 1. When two drugs were implicated the one held responsible was the drug that had highest score with the International Consensus Criteria, usually due to its known hepatoxicity.

Figure 3.

 The most common drugs associated with drug-induced liver injury.

Table 1.   Drugs associated with at least two cases of DILI, number of cases for each responsible drug and clinical characteristics associated with the adverse drug reactions
DrugNumber of casesGenderHCCSMixedDuration of treatment, mean (range)Time to biochemical resolution, mean (range)
  1. M, male; F, female; HC, hepatocellular type of DILI; CS, cholestatic type of DILI; DILI, drug-induced liver injury.

Diclofenac147F/7M75224 (2–60)52 (10–150)
Flucloxacillin82F/6M07110 (2–21)120 (2–540)
Azathioprine54F/1M30298 (30–370)22 (7–30)
Atorvastatin42F/2M211159 (56–370)34 (11–60)
Ciprofloxacin42F/2M2117 (2–7)32 (26–45)
Macrolides33F/0M2017 (4–10)53 (45–60)
Nitrofurantoin22F/0M110189 (7–370)120 (60–180)
Clindamycin21F/1M20021 (7–35)97 (14–180)
Disulfiram21F/1M20056 (42–70)75 (60–90)

Median duration of treatment before diagnosis of liver injury was 30 days (7–68) and the median time to biochemical resolution following discontinuation of the offending drug was 60 days (26–68). The duration of treatment before the diagnosis of DILI and the time to biochemical resolution of the liver test abnormalities in the different types of DILI are shown in Table 2. The time to biochemical resolution was shorter in mixed drug reactions vs. those with a HC type of pattern (Kruskas–Wallis, tied P-value of 0.0069). There was no correlation between the serum bilirubin and the time to biochemical resolution, nor between jaundiced and non-jaundiced patients (data not shown).

Table 2.   The duration of treatment and the time to biochemical resolution for the different types of drug-induced liver injury (DILI)
TypeDuration of treatment, median (IQR; days)Time to biochemical resolution, median (IQR)
Hepatocellular42 (7–72)60 (30–60)
Cholestatic15 (8–83)60 (35–120)
Mixed19 (5–30)21 (10–30)

Table 3 demonstrates the demographics, the duration of treatment, liver tests at presentation and time to biochemical resolution among patients with jaundice. Jaundice resolved in all cases. A wide variety of other drugs were implicated in the other cases, which are shown in Table 4. A total of 43 different drugs were involved in the DILI-associated cases (Tables 1, 3 and 4).

Table 3.   Drugs responsible of DILI associated with jaundice (2 ULN of serum bilirubin) and the nature of the injury
DrugAge (years)GenderDuration of treatmentBilirubin ULNAST ULNALT ULNALP ULNTime to biochemical resolution
  1. ULN, upper limit of normal; AST, aspartate aminotransferase; ALT, alanine aminotransferase; ALP, alkaline phosphatase; DILI, drug-induced liver injury.

Amoxicillin/clavulanic acid65Male103.65.313.88.054
Diclofenac41FemaleOn demand12.828.612.01.221
Omeprazol83MaleOn demand5.
Table 4.   Drugs responsible of DILI but without jaundice (<2 ULN of serum bilirubin) and the nature of the injury
DrugAge (years)GenderDuration of treatmentHC/CS/mixedTime to biochemical resolution
  1. * Medicinal herbs or homeopathic medications.

  2. ULN, upper limit of normal; HC, hepatocellular; CS, cholestatic; DILI, drug-induced liver injury.

Diclofenac/misoprostol53MaleOn demandCS
Octreotide44FemaleOn demandHC30
Aspirin51MaleOn demandHC60

Other liver disease was found in 11 cases (14%). In six of these cases, azathioprine or mercaptopurin (6-mercaptopurin) was the cause of the DILI. The most commonly encountered other liver disease was AIH with six cases, NASH in two cases, alcoholic liver disease, cryptogenic cirrhosis, PBC and PSC in the other cases. Liver biopsy was performed only in 10 cases (12.9%) with findings that were compatible with DILI and in no case did the biopsy findings change the clinical assessment of the liver injury.


A wide variety of different drugs were found to be associated with DILI. A total of nine different drugs with at least two DILI cases each were responsible for 57% of the adverse liver reactions. This is in line with a recent study from our group on DILI associated with fatal outcome in Sweden, showing that the top 10 drugs were responsible for 60% of all fatalities.17 The top nine drugs in the current survey have all been well characterized leading to DILI in susceptible individuals.4, 18

Among patients seen at the out-patient clinic, the most common diagnoses were ALD, NAFLD, HCV, PBC and AIH. Next to these commonly encountered diagnoses, DILI was the most common diagnosis among the patients seen for the first time, representing 6.6% of all patients. Approximately half of the DILI cases were referred for evaluation of liver tests whereas the other half had a follow-up after hospitalization. It might potentially create a bias in our prevalence estimation of the proportion of DILI cases by identifying patients from two different settings, i.e. from the diagnoses database and the voluntary reporting system in our hospital. However, the vast majority of the reported cases were from our unit and had been seen in and reported from the out-patient clinic. This proportion of patients with DILI in the current study was substantially higher than the 0.8% of all hepatology patients that were diagnosed with DILI in Detroit, Michigan.13 As we used an objective instrument in the form of the International Consensus Criteria15, 16 to assess the likelihood of a causality, it is unlikely that our higher figure is due to over diagnosis of DILI. However, it is not clear from their report if the investigators in the US study reviewed all the 4039 medical records in order to look for evidence for DILI.13 The underreporting of DILI is well known19, 20 and the diagnosis of DILI can be obscured by both doctors practice behaviour and difficulties determining the timing of the suspected drug ingestion.2 In fact, it has been demonstrated in a study of computerized data files of health maintenance organization in the US, which doctors often discontinued the suspected drugs, but did not ascribe the liver test abnormalities to drugs in the medical records.2 Similarly, in a study of DILI in hospitalized patients in Switzerland, in less than half of cases considered by the authors to be DILI, liver injury was mentioned among diagnoses or in the doctor's discharge letter.21 Thus, evaluation only of cases with a firm diagnosis of DILI in a computerized diagnoses database may lead to a severe underestimation of the problem of DILI. In the present study, all medical records of patients with a diagnosis of liver disease were scrutinized in order to find evidence for the occurrence of DILI. All the cases with a clinical suspicion of DILI were systematically assessed with the International Consensus Criteria15, 16 leading to an exclusion of approximately 30% of cases initially classified as DILI. This is in line with our previous study showing that approximately 30% of cases reported to SADRAC suspected to have caused fatal DILI, did not fulfil the International Consensus Criteria for a possible causal relationship with a drug. In another study from UK, drug-induced liver disease was wrongly implicated in almost half of alleged cases of DILI when subjected to critical review.8 Thus, performing a strict causality assessment is of crucial importance in order to achieve a high diagnostic accuracy. Although the International Consensus Criteria have some limitations, they are still widely used and accepted to be the best available diagnostic tool in DILI.22–24

The strengths of the current study is the thorough analysis of all medical records of patients with a diagnosis of liver disease looking for evidence of DILI and thereafter performing a strict causality assessment of all cases. However, the weakness is its retrospective nature. There were also some limitations in the diagnostic work-up and some of the cases were not followed up to complete normalization of the liver tests but only to the point where it was not considered clinically motivated to follow the development of liver tests after dechallenge of the offending drug.

Our calculation of the crude incidence rate of 2.3 per 100 000 inhabitants/year with a DILI is remarkably similar to the 2.4 per 100 000 inhabitants/year reported from a population-based case–control study using the UK-based General Practice Research database.3 However, a careful prospective survey of DILI in the general population in France revealed a much higher incidence of DILI.12 Sgro et al.12 found the incidence of DILI to be approximately 14 per 100 000 and year and spontaneous reporting to the French regulatory authorities was at the most, 1/16 of the true reporting. The crude incidence of DILI in our study is likely to be an underestimation as we only searched for patients seen for the first time in our out-patient clinic, half of whom had been hospitalized for their DILI before they came to the out-patient clinic. Thus, a number of in-patients not followed up as out-patients could have been missed. Furthermore, a probably small number of patients with severe DILI may have been missed in the present study, as such patients may have been admitted to the intensive care unit, and if fatal or transplanted never been registered to the out-patient clinic register. On the other hand, the figures by Sgro et al.12 could to some extent reflect the fact that the prescription rate in France is at least twice as high in that country when compared with Sweden.19

In agreement with other studies, the majority of patients with at least a causal relationship with DILI were older than 50 years (64%), similar to patients in the general population12, 25 and among DILI cases reported to the authorities.9, 10 Old age has been observed in association with adverse hepatic reactions to drug in general.26 Apart from frequent use of drugs among older patients it has been increasingly suggested that hepatic adverse reactions are age-related27 and old age has been shown to increase the risk for halothane-, isoniazid-, nitrofurantoin- and flucloxacillin-induced hepatitis.4, 18, 28

In the present study, 56% of the DILI cases were females. Drug-induced liver injury has been reported to occur more often in females than in male patients in some6, 9, 11, 12 but not in other studies.13, 25, 29, 30 However, Sgro et al. observed similar DILI incidence rates in the two sexes before the age of 49 years, whereas DILI became more than twice as high in women after that age, suggesting an increased risk of DILI associated with menopause.12 Other studies have shown conflicting results concerning the gender-associated risks for DILI with both a higher frequency in females11 as well in males12 who are older than 50 years of age. In the present study, we found a similar frequency in females (65%) above 50 years as in males (62%) in this age category. A somewhat unexpected finding was the significantly higher occurrence of HC DILI among females in comparison with males. This was mostly due to a high frequency of HC damage in females younger than 50 years, which in 80% of cases had HC type of DILI whereas among females older than 50 years, only 53% of cases had HC injury. It was a little surprising to find significantly more males to have a CS/mixed pattern and we do not have any explanation for this finding. We found the similar proportion of NSAID-associated DILI in the two sexes. Other studies have showed a significant association between liver disturbances and NSAID use in women but not in males.31, 32 Our finding of an increased occurrence of HC pattern among females is unclear but is in line with the observation that 75% of the patients with HC injury on the liver transplant list because of non-acetaminophen DILI in the US were women.33

None of the patients in the current study died from their DILI, in spite of the fact that 29 patients (38%) had jaundice, 12 of whom fulfilled the Hy's rule criteria with a HC type of reaction and ALT ≥3 ULN, a condition usually associated with approximately 10% mortality.6, 25 We have also demonstrated a rather high mortality rate (7.8%) in CS type of injury, with concomitant jaundice, and similarly high ALT values were found in 14 of our patients in the present study.6 The reason why we did not see any fatal cases in the present study may be attributable to the fact that the most severely ill patients never came to the out-patient clinic but were admitted directly to the wards.

Limited data exist on the time to biochemical resolution of liver tests abnormalities in DILI. In the prospective French survey, a mean time to resolution was approximately 30 days whereas approximately 3 months have been reported in other studies.13, 25 We observed a mean time of approximately 2 months to resolution among patients followed to normalization of liver tests but in several cases the patients were not followed to complete normalization of liver tests. As this was a retrospective study, it was not possible to determine whether any case of DILI has led to chronic liver injury. In a prospective study on DILI from Spain, approximately 10% of patients had liver abnormalities at follow-up more than 3 months after the original insult.25 In fact, the only published study on long-term outcome of DILI is a retrospective study in patients identified by a histological database, demonstrating a considerable number of patients with significant abnormalities in liver tests or on imaging at follow-up.34 More studies on the long-term effects on DILI are clearly needed.

The current study revealed that the most common type of drugs were antibiotics accounting for 30% of all cases, which is very similar to the approximately one-third of patients with DILI and concomitant jaundice6 as well as in DILI associated with fatal outcome17 that we have observed in previous studies. Other investigators have shown similar high frequency of antibiotics with suspected DILI.12, 13, 21, 25, 34 Obviously, local therapy strategies as well as marketing of different antibiotics in different countries might explain the differences in the published data on which different antibiotics are most frequently associated with DILI. In the present study, flucloxacillin, which has a well-documented hepatotoxicity,28 was the most common antibiotic, with a prolonged mean time to resolution. Flucloxacillin was also highly prevalent in patients with suspected DILI identified by the UK general practice database.3 However, among patients in the UK study,3 amoxicillin/clavulanic acid was the most commonly encountered antibiotic as well in the prospective study from Spain25 whereas we only found one case in our study. This probably reflects the very limited use of this drug combination in Sweden when compared with some other countries.3, 25

Although the risks of hepatoxicity because of NSAIDs are low, DILI associated with NSAIDs has increasingly,6, 25, 29, 30 but not consistently12, 13 being recognized, probably because of the very prevalent use of these drugs among the general population. Among the NSAIDs, diclofenac has been shown to induce DILI in a small minority of users35 with 8% fatality rate in diclofenac-associated hepatitis in the US.36 In fact, we recently observed 20% mortality among diclofenac-associated DILI in patients with concomitant jaundice6 and a total of 56 cases with fatal outcome had been reported to WHO Collaborating Center for International Drug Monitoring with diclofenac suspected DILI in a recent survey.29 The current study revealed that diclofenac was the single agent most often responsible for DILI, in total accounting for 18% of all cases.

We also observed the absence of DILI in many very commonly used drugs, such as antidepressants (except one case), diuretics, low-dose aspirin, calcium inhibitors and β-blockers. Only two statins were suspected to have caused DILI in the current study, atorvastatin and fluvastatin which both have been reported to be associated with DILI.12, 37 We found medicinal herbs to be responsible for 5% of DILI cases which is somewhat higher than the 2% of cases collected prospectively in Spain.25

In conclusion, DILI is a significant clinical problem among hepatology referrals with diagnostic challenges. Antibiotics and diclofenac were the most common causes of DILI among out-patients.


The study was supported by the Faculty of Medicine, University of Gothenburg.