Cohort study of hepatotoxicity associated with nimesulide and other non-steroidal anti-inflammatory drugs

Introduction

Nimesulide, a non-steroidal anti-inflammatory drug, is marketed in more than 50 countries. In March 2002, Finland suspended the marketing of nimesulide because of an associated high frequency of hepatotoxicity.1 Spain followed in May 2002, but not other European countries, such as Italy and France.2^–5 Nimesulide is the most prescribed non-steroidal anti-inflammatory drug in Italy and Portugal, with Italy accounting for half the worldwide market. We compared the incidence of acute hepatotoxicity associated with nimesulide and other non-steroidal anti-inflammatory drugs.

Methods

During 1997-2001 we carried out a retrospective cohort study in the Umbria region of Italy, with a population of around 835 000. Patients were enrolled if they had received at least one prescription for a nonsteroidal anti-inflammatory drug within the national health service between 1 January 1997 and 31 December 2001. (The monitoring system that provided the information on use included only prescriptions issued within the health service; private purchase is therefore not captured in the cohort). In Italy most commonly prescribed non-steroidal anti-inflammatory drugs are also available as over the counter products—for example, diclofenac, ketoprofen, piroxicam, naproxen, ibuprofen—but prescription only non-steroidal anti-inflammatory drugs are often purchased privately.

Drug use was defined as current according to the duration of the prescription (in defined daily doses) plus two weeks; recent for the 90 days after the end of the current period; and past use for the days up to 12 months after the prescription date. If a new prescription of the same drug occurred during the current period, time of use continued to accumulate in the particular category. When different non-steroidal anti-inflammatory drugs were dispensed in consecutive overlapping intervals of current use, time after the new prescription was allocated to the mixed use category.

Potential cases were all admissions to hospital for acute non-viral hepatitis. Codes were retrieved for main or secondary diagnoses according to ICD-9 (international classification of diseases, 9th revision) from the regional archive of hospital discharges: hepatitis, unspecified (573.3), acute and subacute necrosis of liver (570), other specified disorders of biliary tract (576.8), other specified disorders of liver (573.8), unspecified disorders of liver (573.9). No information was available on patients admitted to hospital outside the Umbria region.

Information on medical history, concomitant clinical conditions, drug use during the six weeks before admission, and laboratory data were collected from the clinical records, blinded to exposure status. Patients were excluded if they had malignancies, viral or chronic hepatitis, lithiasis of common biliary tract, a cholecystectomy performed during the admission to hospital included in the study, or normal liver function.

Abnormal liver function was defined as any increase between the upper limit of normal range and twice the upper limit of normal for alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase, or total or conjugated bilirubin. Liver injury was defined as the increase over twice the upper limit of normal for alanine aminotransferase or conjugated bilirubin, or a combined increase of aspartate aminotransferase, alkaline phosphatase, and total bilirubin, provided one of them was above twice the upper limit of normal.6^–8 To also include more severe cases we restricted the analysis to increases over five times the upper limit of normal. Information on liver tests was available for all cases.

A nested case-control study was carried out to control for potential confounders: use of any other drug (different from non-steroidal anti-inflammatory drugs) during the 30 days before admission and number of prescriptions for non-steroidal anti-inflammatory drugs during the current period (one and more than one packet). For each case we randomly selected 20 controls from the cohort, matched for sex and age (difference of one year). The date of admission for cases was the index date for the matched controls. Use of non-steroidal anti-inflammatory drugs at the index date for both cases and controls was defined as current, recent, or past, as in the cohort analysis.

Incidence was calculated by dividing the number of cases in each category of use by the corresponding person years of non-steroidal anti-inflammatory drug use. Poisson regression was used to control for the confounding effect of age and sex in the cohort study. Conditional logistic regression was used for the analysis of the case-control study. The analysis was performed with SAS (version 8.00) and SPSS (version 10.0.7) software.

Results

Between 1997 and 2001 around 2 million prescriptions for non-steroidal anti-inflammatory drugs were issued within the national health service in Italy and were included in our study. Nimesulide was the most prescribed drug (551 000 prescriptions). The characteristics of users of the different non-steroidal anti-inflammatory drugs were similar (table 1). Most of the prescriptions were related to short term treatment (1.2 packets per prescription for nimesulide and 1.3 for other non-steroidal anti-inflammatory drugs).

The 397 537 participants who received at least one prescription for a non-steroidal anti-inflammatory drug contributed to more than 770 000 person years (current use, 141 000 person years; recent use, 254 000 person years; past use, 378 000 person years; see table 1).

We excluded 168 of 819 (20.5%) potential cases because of hepatopathy as a secondary diagnosis and concomitant malignancy (figure). Of the remaining 651 clinical records, 568 (87.3%) were retrieved. Non-retrieved potential cases (12.7%) were proportionally distributed over the study drugs. In total, 392 patients were excluded.

Flow of participants through trial

• Download figure
• Open in new tab
• Download powerpoint

Forty two of the 176 cases of hepatopathy included in the final analysis occurred during current use of a non-steroidal anti-inflammatory drug (incidence 29.8 per 100 000 person years). Compared with the incidence for past use (18.2), we estimated a rate ratio (adjusted for age and sex) of 1.4 (95% confidence interval 1.0 to 2.1). This ratio increased among elderly participants: after taking into account the effect of sex and current use of non-steroidal anti-inflammatory drugs, those aged over 75 had a 5.7-fold increase in the risk of hepatopathy compared with people under 45. The risk of hepatopathy in males was increased by 1.5 (1.2 to 2.2) compared with females (table 2).

The risk of hepatopathy among current users of nimesulide was slightly higher than for other non-steroidal anti-inflammatory drugs (rate ratio 1.3, 0.7 to 2.3). When the analysis was restricted to patients with liver injury this ratio was 1.7 (0.9 to 3.3). When only more severe diagnoses were included in the analysis, the rate ratio among users of nimesulide was 1.9 (1.1 to 3.8; table 3). For use of individual non-steroidal anti-inflammatory drugs compared with past use, the increase in the incidence of liver injury was 2.2 (1.3 to 3.9) for nimesulide and 1.5 (0.7 to 3.2) for diclofenac, the second most prescribed nonsteroidal anti-inflammatory drug (table 4). The number of events for individual non-steroidal anti-inflammatory drugs was too small to allow for comparison of risks.

The analysis carried out in the nested case-control study confirmed the risk estimates of the cohort analysis. The odds ratio of liver injury developing among users of nimesulide compared with users of other nonsteroidal anti-inflammatory drugs was 1.9 (0.6 to 5.6), after controlling for prescriptions of other drugs received in the 30 days before admission to hospital (index date) and for number of prescriptions for nonsteroidal anti-inflammatory drugs. Those who received drugs during the 30 days had a twofold increased risk (adjusted odds ratio 2.0, 0.6 to 6.6). No effect was associated with number of prescriptions for non-steroidal anti-inflammatory drugs during the current period (0.7, 0.3 to 1.9).

Among current users, the median interval between prescription for a non-steroidal anti-inflammatory drug and admission to hospital was 11 days, and only two of the 42 patients took the drugs for more than 60 days. No fulminant hepatitis was observed. Among patients with liver injury a trend towards the normalisation of liver function (a positive dechallenge) could be assessed in 12 users of nimesulide (75%) and in 13 users of other non-steroidal anti-inflammatory drugs (65%). In the remaining patients there was either no follow up information or the length of stay was too small for comparisons (inconclusive dechallenge) (table 5).

Some patients admitted to hospital for liver injury received one or more prescriptions for the same drug after discharge: eight (50%) for current users of nimesulide (either alone or mixed use) and five (25%) for current users of other non-steroidal anti-inflammatory drugs. No participants were readmitted to hospital.

Three deaths occurred among current users (diclofenac, nimesulide, and piroxicam). Cause of death was considered unrelated to liver injury: one patient had normalisation of liver function and the other two had cardiac failure, but no worsening of hepatopathy.

Discussion

The risk of hepatopathy among patients taking non-steroidal anti-inflammatory drug is small. Despite its withdrawal in Finland and Spain because of reported hepatotoxicity, nimesulide was associated with only a small increased risk. The main risk factor for hepatotoxicity is age.

One strength of our study was that we followed a general population, in which almost 2 million prescriptions for non-steroidal anti-inflammatory drugs were issued during a five year period. Limitations of our study were that drug use was estimated through a prescription monitoring system, and no information was available on the actual use of the drugs, such as indications and dosage. A proxy for this information was the number of packets prescribed and the defined daily dose for each user. It seems that both nimesulide and other non-steroidal anti-inflammatory drugs are used as short term treatments. It is possible that confounding by indication may play a part in our study. The indications for prescription within the Italian national health service are the same for nimesulide and other non-steroidal anti-inflammatory drugs. Moreover, the users of both nimesulide and other non-steroidal anti-inflammatory drugs were similar, inferred from age, sex, expected duration of prescription, and concomitant prescriptions.

Our findings are consistent with available evidence. A cohort study in Canada concerned 1.5 million prescriptions for non-steroidal anti-inflammatory drugs.9 In our study, 16 current users of non-steroidal anti-inflammatory drugs were admitted to hospital for acute liver injury, which corresponds to a rate of 9 per 100 000 person years and 1 per 100 000 prescriptions. We found a high incidence of hepatopathy, but if the number of prescriptions is taken as the denominator, the incidence of liver injury is 1.7 per 100 000 prescriptions, similar to the Canadian study.

The largest study to date is based on 2.1 million prescriptions for non-steroidal anti-inflammatory drugs from British general practitioners.10 Out of the 23 patients with acute liver injury (1.1 per 100 000 prescriptions), eight were admitted to hospital (0.4 per 100 000). The age range of the 23 cases was between 26 and 80, whereas the age range of our patients was between 27 and 96, with almost 50% of the patients older than 80.

In Finland, spontaneous reporting of adverse effects associated with nimesulide reached more than 100 per 100 000 person years. The corresponding incidence for other non-steroidal anti-inflammatory drugs was less than 1 per 100 000 person years. In Spain, nimesulide was associated with the highest spontaneous reporting rate for hepatopathy (9.37 cases per million packets).3 A large variability in reporting rates for other non-steroidal anti-inflammatory drugs has been observed, which is often the case with spontaneous reporting systems.

We found only a small increase in the risk associated with nimesulide, consistent with observations in other countries, such as Italy, France, and Portugal. In these countries the reporting of hepatotoxicity was similar for nimesulide and other non-steroidal anti-inflammatory drugs.

The potential mechanism of hepatotoxicity associated with nimesulide is unknown. It has been suggested that symptomatic hepatic effects of most non-steroidal anti-inflammatory drugs are usually mild.11 In most of the patients in our study admitted to hospital because of liver injury, there was evidence of a trend towards the normalisation of liver function. As with two other studies, no fulminant hepatitis was observed.4 5 This suggests that some of the spontaneously reported cases of fulminant hepatitis attributed to non-steroidal anti-inflammatory drugs are linked to concomitant conditions that would lead to exclusions in epidemiological studies.

A final consideration is the need to evaluate the overall risk profile of non-steroidal anti-inflammatory drugs, particularly the risk of serious gastroduodenal complications (bleeding and perforation), which are almost 10 times higher than hepatotoxicity. The incidence of admission to hospital for bleeding or perforation among non-users of non-steroidal anti-inflammatory drugs is around 100 events per 100 000 persons per year. Among users of non-steroidal anti-inflammatory drugs, assuming a relative risk of 4, around 400 events per 100 000 persons per year are expected, with around 300 extra cases per 100 000 persons per year.12 13 In three observational studies in Italy, nimesulide was generally of average to low risk and other non-steroidal anti-inflammatory drugs, such as ketorolac and piroxicam, were more gastrotoxic.14^–16

Our study was coordinated by the Istituto Superiore di Sanità, the scientific body of the Italian National Health Service. We thank Paolo Di Loreto, Barbara Gamboni, Carlo Romagnoli, and Mariangela Rossi (Umbria Regional Health Authority) for providing the information on hospital admissions and drug use in Umbria; the NHS staff who helped to retrieve the clinical records; Antonella Germani for reviewing the clinical records; Bruno Caffari, Marina Maggini, Alfonso Mele, Roberto Raschetti, Stefania Spila Alegiani, Tommaso Stroffolini, Nicola Vanacore (Istituto Superiore di Sanità), Achille Caputi (University of Messina), and Albano Del Favero (University of Perugia) who participated in the discussion and review of the study protocol and the study findings; and Sara Modigliani for editorial support.