Nimesulide-induced hepatotoxicity: A systematic review and meta-analysis

Objective This study aimed to evaluate the risk for hepatotoxicity with nimesulide, a non-steroidal anti-inflammatory drug (NSAID) available in Republic of Korea but withdrawn from the market in several countries. Methods A systematic review and meta-analysis were conducted of studies retrieved from PubMed, EMBASE, Cochrane, the Research Information Sharing Service and ClinicalTrials.gov up to September 2017. All studies reporting nimesulide-associated hepatotoxicity in patients as compared with the unexposed or the exposed to other NSAIDs were included. Studies using spontaneous reporting databases were included to estimate reporting odds ratio (ROR) of hepatotoxicity associated with nimesulide exposure. The association between nimesulide use and hepatotoxicity was estimated using relative risk (RR) and ROR with 95% confidence interval (CI). Results A total of 25 observational studies were eligible for review. In a meta-analysis of five observational studies, nimesulide was significantly associated with hepatotoxicity [RR 2.21, 95% CI 1.72–2.83]. From studies using spontaneous reporting databases (n = 6), rates of reported hepatotoxicity were significantly higher in patients using nimesulide, compared with those treated with other NSAIDs [pooled ROR 3.99, 95% CI 2.86–5.57]. Of a total of 33 patients from case studies and series, the majority (n = 28, 84.8%) were female, and the mean age (± standard deviation) was 56.8 (± 15.6) years. Almost half of the patients on nimesulide (45.5%) either required liver transplantation or died due to fulminant hepatic failure, of whom a third developed hepatotoxicity within less than 15 days of nimesulide administration. Conclusions Our study findings support previous reports of an increased risk for hepatotoxicity with nimesulide use and add to existing literature by providing risk estimates for nimesulide-associated hepatotoxicity. As the limited number of studies with primarily observational study designs were included in the analysis, more studies are needed to further describe the effects of dose and length of treatment on the risk for hepatotoxicity.


Introduction
Nimesulide is a non-steroidal anti-inflammatory drug (NSAID) with preferential inhibitory activity on cyclooxygenase 2 (COX-2) enzyme [1]. The drug was first launched in Italy in 1985 and was subsequently marketed in more than 50 countries, including South Korea [2]. It has potent analgesic, anti-inflammatory, and antipyretic properties, with a relatively low risk for gastrointestinal side effects, as demonstrated by numerous clinical trials [3,4]. Moreover, nimesulide, when administered orally, is rapidly and extensively absorbed, thus allowing effective pain control [3,5]. However, nimesulide induced hepatotoxicity was first reported in 1997 [6] and severe, and even fatal, cases of liver injury have been reported in patients who received nimesulide treatment [7]. Consequently, the use of nimesulide was restricted or withdrawn from the market in 2002 in Spain and Finland, followed by several other countries [8].
A number of observational studies that evaluated the safety profile of nimesulide were published [8][9][10][11][12] at around the time when nimesulide was initially banned in several countries. In 2004, the European Medicines Agency (EMA) recommended a restriction of nimesulide indications, as well as its maximal daily dose [13]. However, in May 2007, the Irish Medicines Board, the former regulatory agency of the Health Products Regulatory Authority, announced the marketing suspension of oral nimesulide-containing products due to a number of cases of fulminant hepatic failure requiring liver transplantation [14]. This prompted the EMA to undertake a further safety review of the drug, which, on completion in 2012, led the agency to support the continuous use of nimesulide, based on drug benefits outweighing the risks for liver toxicity [15]. However, this decision met with disagreement among some members of the Committee for Medicinal Products for Human Use within the EMA [15]. Subsequent widespread controversy surrounding the safety of nimesulide has led to varying regulatory decisions on restricting its use across different European countries.
To our knowledge, to date, there are no published studies using systematic evaluation methods to quantitatively assess the safety profile of nimesulide related to hepatotoxicity in peerreviewed journals. The aim of this study was to evaluate hepatotoxic effects induced by nimesulide. We conducted a systematic review of the published literature, including case reports and series, on hepatotoxicity associated with the use of nimesulide in human patients and performed a meta-analysis of studies that assessed any hepatic adverse event outcomes.

Search strategy and data sources
A systematic review of the literature was performed in accordance with the Preferred Reporting of Systematic Reviews and Meta-Analyses (PRISMA) guideline (S1 Checklist) [16], using the following databases for studies published within the specified periods: PubMed (July 1998 to September 2017), Embase (August 1998 to September 2017), the Cochrane Central Register of Controlled Trials (November 1999 to September 2017), and the Research Information Sharing Service (Korean bibliographic database; April 1988 to September 2017). In addition, search in ClinicalTrials.gov database was also conducted to include unpublished trials. The following keywords were used to identify relevant articles on nimesulide-induced hepatotoxicity: 'liver toxicity', 'hepatotoxicity', 'chemical and drug-induced liver injury', 'drug-induced hepatitis', and 'nimesulide' (S1 Table). No restriction was imposed in terms of study design and publication language. Additionally, the reference lists of retrieved articles were also manually searched.

Study selection
Study designs for the selection included randomized controlled trials, observational studies, case reports, and case series. Hepatotoxicity outcomes were identified in patients exposed to nimesulide as compared with unexposed patients or those with exposure to other NSAIDs. Study participants were of either sex and any age. Of the identified articles, duplicates were removed using the bibliographic software EndNote X8.1 (Thomson Reuters, Philadelphia, PA, USA). One author (JK) identified potentially relevant articles for inclusion by titles and abstracts, while two authors (JK/SK) independently reviewed the entire manuscripts. Any disagreements between the authors were resolved by discussion or by a third author (EL). Studies were considered eligible for inclusion if they described hepatotoxicity associated specifically with the use of nimesulide. Exclusion criteria were: (1) non-human studies, (2) non-original research article type, (3) cases with pre-existing liver disease, and (4) same data also reported in another study.

Data extraction and quality assessment
We extracted information from each study, including study design, source of data, population characteristics, and study outcomes. Additional information from case series and case reports were gathered on the duration of nimesulide treatment prior to initial presentation of signs and symptoms of hepatotoxicity, concurrent medications, clinical features, evidence of hypersensitivity or autoimmune reaction, and laboratory results on admission. Where available, we recorded the odds ratios (ORs) and relative risks (RRs), as well as the proportion of patients who experienced liver injury that either was reported or could be calculated.
Since all included studies were non-randomized, we used the Newcastle-Ottawa Scale (NOS) to assess the quality of observational studies, except for case series or case reports [17]. The NOS uses a star system to assess the quality of a study based on three domains: selection, comparability, and outcome (cohort studies) or exposure (case-control studies), with the quality of the study rated as low (0-3 stars), medium (4-6 stars), or high (7-9 stars). Two authors (JK/SK) evaluated the quality of the studies, and any discrepancy was resolved by consensus reached including the third author (EL).
For quality assessment of case series and case reports, the Roussel Uclaf Causality Assessment Method (RUCAM) [18] was used to quantify the strength of the association between liver injury and use of nimesulide. Causality was classified as: highly probable, probable, possible, and unlikely. The RUCAM provides different subscales, depending on the pattern types of liver damage which are classified as hepatocellular, cholestatic, and mixed liver injury [18]. These three types of liver damage can be differentiated using the R value calculated as the alanine aminotransferase (ALT)/alkaline phosphatase (ALP) activity measured at the time liver injury is suspected, with both activities expressed as multiples of the upper limit of normal [18]. The hepatocellular pattern of liver damage was defined as R values �5, mixed pattern as R values >2 and <5, and cholestatic pattern as R values �2 [18]. If a study did not report the type of liver injury, we calculated the R value to determine the type of liver damage.

Statistical analysis
The primary analysis focused on assessing the risk for hepatotoxicity among patients treated with nimesulide. We used the ORs (adjusted, when available) or rate ratios reported in the case-control or cohort studies, respectively; we calculated the ORs or rate ratios, if not reported, using the proportion of patients reported with nimesulide-induced hepatotoxicity in each study. The association between nimesulide use and the risk for hepatotoxicity was estimated using relative risks (RRs) as well as 95% confidence intervals (CIs).
For studies using spontaneous reporting databases, we conducted a comprehensive disproportionality analysis by applying a case/non-case method. Cases included all studies reporting hepatotoxicity, whereas non-cases included all other reports recorded during the study period. Association between hepatotoxicity and use of nimesulide was estimated using reporting odds ratio (ROR) as a measure of disproportionality. The ROR is the ratio of the odds of nimesulide exposure among cases to the odds of nimesulide exposure among non-cases [19]. We pooled raw data of the proportion of reports for each NSAID, including nimesulide, from studies to compute the RORs, compared with other NSAIDs, and their corresponding 95% CIs.
The meta-analysis was conducted separately, depending on whether the measure of risk estimate was the RR or ROR. Sensitivity analyses were carried out to explain possible heterogeneity between studies by including studies in, or excluding them from, the meta-analysis, based on the study design and measure of the RR (ie., OR and rate ratio). Statistical heterogeneity across studies was assessed using the I 2 statistic and Cochran's Q test. An I 2 value of �50% or a Cochran Q test P value of <0.10 indicated significant heterogeneity [20]. Overall estimate of the RR was obtained from a random effects model when statistic heterogeneity was present; otherwise, a fixed effects model was used. Data analyses were performed using SPSS version 23.0 (IBM SPSS Corp, Chicago, IL, USA) and Comprehensive Meta Analysis version 2.2 (Biostat, Englewood, NJ, USA).

Search results
A total of 265 potentially eligible articles were identified by searching the three electronic databases using the keywords, as well as the relevant reference sections. Of these, 60 duplicate records were identified and removed. After screening the article abstracts and titles, 163 articles were excluded, and the remaining 42 articles underwent detailed full-text evaluation. Finally, 25 studies including 2 cohort studies, 1 case-control, 1 case-crossover study, 5 case/noncase studies, 3 case-population studies, 4 case series and 9 single-case reports were eligible for inclusion ( Fig 1) and are summarized in Table 1. One [21] of the two listed cohort studies was analyzed as a case series study because only the number of cases and case descriptions were provided, without a clear measure of association. One [22] of the nine single-case reports performed an analysis generating the ROR using the World Health Organization Uppsala Monitoring Centre (WHO/UMC) pharmacovigilance database and was pooled with the studies using spontaneous reporting databases in our study. Of the included studies, 11 studies were included in quantitative analysis.

Study characteristics and quality
The majority of observational studies included in our review were conducted in a number of European countries, including Italy, Spain, Switzerland, Ireland, France, Greece, the Netherlands, Portugal, and the UK, as well as in one Asian country, namely Taiwan ( Table 1). Four of the studies identified patients from hospital medical records; five used administrative pharmacovigilance databases; two used both medical records and national health insurance databases, and one study used data from liver transplantation centers. In quality assessment, three studies were found to be of high quality, four studies of medium quality, and six of low quality (S2 and S3 Tables).
We identified 33 patients who were reported to have nimesulide-induced hepatic injury from the case reports and case series included in the study (Table 2). Cases were reported from 12 countries, including Israel, Belgium, France, Greece, Italy, Ireland, Iceland, Spain, Switzerland, Serbia, Singapore, and South Korea. The mean age (± standard deviation) of the patients was 56.8 ± 15.6 years (median 57 years; range 18-81 years). Age of �55 years was a risk factor found in 22 (66.7%) patients. The majority of patients with liver injury were female (n = 28, 84.8%), and the dose of nimesulide reported in the studies was either 100 mg or 200 mg daily, except for two patients who were given 150 mg or 600 mg daily, respectively, as well as an unreported dose for three patients. The duration of nimesulide treatment prior to initial presentation of signs and symptoms of hepatotoxicity ranged from 8 hours to 189 days (median 42 days).
Out of 33 identified patients, only two patients [29] had signs of hypersensitivity such as an increased eosinophilia or liver specific autoantibodies. The type of liver injury reported was hepatocellular in 14 cases (42.4%), cholestatic in six cases (18.2%), and mixed in four cases (12.1%), whereas the type was unknown in nine cases (27.3%) due to insufficient data. Eighteen (54.5%) of the patients recovered; of the remaining 15 (45.5%) patients who underwent   (7) ALT, alanine aminotransferase; AP, alkaline phosphatase; AST, aspartate aminotransferase; F, female; M, male; ×ULN, multiples of the upper limit of normal.
a The duration of nimesulide treatment prior to initial presentation of signs and symptoms of hepatotoxicity.
b Causality was assessed using the Roussel Uclaf Causality Assessment Method (RUCAM) and divided into four categories: highly probable, probable, possible, and unlikely.
c Nimesulide was discontinued due to hepatotoxicity and then it was rechallenged after 2 months. liver transplantation, nine survived and six died. Moreover, 5 out of these 15 patients developed hepatotoxicity within less than 15 days of nimesulide administration. Using the RUCAM scoring system, nimesulide-attributable hepatotoxicity was probable in 18 cases, highly probable in 11 cases, and possible in four cases.

Meta-analysis
The findings of the meta-analysis are summarized in Figs 2 and 3. Five studies provided data suitable for analysis of hepatotoxicity outcomes. Use of nimesulide significantly increased the risk for hepatotoxicity (RR 2.21, 95% CI 1.72-2.83) (Fig 2). A fixed effects model was applied because heterogeneity across the studies was not statistically significant (I 2 = 18.8%, P = 0.294). Sensitivity analysis showed no substantial change in pooled risk estimates upon exclusion of each of the included studies from the analysis (S4 Table). After excluding studies that reported rate ratios, the two remaining studies showed a statistically significant increased risk (fixed effects RR 2.43, 95% CI 1.82-3.26), with no evidence of heterogeneity (I 2 = 0%, P = 0.474) (S4  Table). Excluding the two case-population studies had no effect on the overall risk for hepatotoxicity (fixed effects RR 2.23, 95% CI 1.76-3.00, I 2 = 0%, P = 0.479) (S4 Table).
A total of six studies provided data suitable for the disproportionality analysis of hepatic adverse events in patients treated with nimesulide, based on the pharmacovigilance databases. Findings showed that use of nimesulide was associated with a significantly greater proportion of reported hepatic adverse events, compared to use of all other NSAIDs (random effects ROR 3.99, 95% CI 2.86-5.57), but the overall estimate was highly heterogeneous (I 2 = 89.9%, P <0.001) (Fig 3). However, excluding the three analyses using the French [23,24] and WHO/ UMC [22] pharmacovigilance databases showed a statistically significant increased risk (fixed effects ROR 6.10, 95% CI 5.55-6.70), with no heterogeneity among the studies (I 2 = 0.0%, P = 0.849) (S4 Table).

Discussion
Findings from this systematic review and meta-analysis indicated that nimesulide was associated with an increased risk for liver injury. While nimesulide-induced hepatotoxicity has been recognized in previously published studies [3,41], we found no scientific reports quantifying the pooled risk, apart from an official report released by the EMA in 2012 that included two studies. We believe that our quantified measures as the pooled RR and ROR can be considered as one of the strengths of our systematic review that included all published studies until late 2017.
ROR measures for the evaluation of nimesulide-associated hepatotoxicity obtained from various databases and the patterns of NSAID use varied among countries, which could, at least in part, explain the differences in reporting rates. It has been previously shown that health care professionals' reporting behavior for adverse drug reactions differed slightly across the European Union [42]. Media attention and publicity resulting in increased reporting, known as notoriety bias [43], could explain the differences in reported RORs in studies originating even from the same country [23,24]. Withdrawal of nimesulide from Spain and Finland in 2002 [8] could also have contributed to the subsequent increased reporting rates in other European countries.
In our analysis of case reports and case series, the majority of cases of nimesulide-associated hepatotoxicity occurred in elderly and female patients. These findings are in agreement with a few published studies on drug-induced liver injury (DILI) related to NSAID use in particular [44][45][46]. The increased incidence of DILI in the elderly carries biologic plausibility in terms of pharmacokinetic changes associated with aging. Conflicting reports, however, on genderrelated occurrence of DILI were also published [23,24]. A retrospective study from Spain showed an overall similar gender distribution in DILI cases [47], whereas a case-control study from France demonstrated a significant association of liver injury caused by NSAIDs in females [46]. Recent studies have shown a relationship between female sex and the hepatocellular pattern of DILI leading to poor outcomes [48]. Findings from our study showed a higher rate of hepatocellular DILI in females, compared to males, with the majority of patients with fatal outcomes directly related to DILI itself or as a result of liver transplantation for DILI being female. Although the same pattern of female preponderance for hepatocellular injury has been reported in various case series [21,31], more epidemiologic assessments using well-validated study designs, as well as pathologic studies, are needed to explain this gender difference and its prevalence as well as the patterns and severity of nimesulide-associated hepatotoxicity.
Our study showed that almost half of patients required liver transplantation or died as a result of fulminant hepatic failure. Of importance, a third of these patients developed hepatotoxicity within less than 15 days of nimesulide administration, which is the maximum duration of nimesulide treatment as approved by the EMA. One study reported that the risk for liver injury increased with treatment duration, even when the treatment period is shorter than 15 days [9]. This results highlighted the needs for closer monitoring from the early phase of the nimesulide use process and healthcare professional should be aware of the nimesulideinduced hepatotoxicity. The majority of cases of DILI are idiosyncratic, occurring in most instances within 5-90 days after ingestion of the causative drug [49]. Similarly, in our study, nimesulide-induced hepatotoxicity generally occurred between 5 and 90 days after initiation of nimesulide treatment, suggesting an idiosyncratic mechanism is likely to be involved. Although the clinical signs of hypersensitivity were not observed in the majority of cases in our analysis, an increased eosinophil was presented in two patients. In addition, some studies suggested that their patients' hepatotoxicity were related to metabolic idiosyncrasy [33]. Therefore, these findings indicated a potential mechanism of nimesulide-induced liver injury involving both immunologic and metabolic pathway. Further research is needed for elucidating biological plausibility of nimesulide-associated hepatotoxicity.
This systematic review has a few limitations. Firstly, only observational studies were included in the analysis, as no randomized controlled trials were available on the risk for hepatotoxicity with nimesulide use. While randomized controlled trials are superior in study design validity, they are usually underpowered when detecting rare events. Therefore, it is often inevitable to rely on observational study designs or secondary data analyses using heterogeneous data sources to evaluate safety outcomes at the expense of strong study validity. Secondly, our study had to apply less stringent inclusion criteria, as few published studies specifically investigated nimesulide-related safety outcomes as their primary research aim. In order to capture all potential adverse effects, our inclusion criteria were not limited to nimesulide-related liver injury as the primary research outcome. Despite our efforts to include as many studies as possible for evaluation, the limited number of studies available precluded any subgroup analysis to examine the effects of age, gender, dose, and length of treatment on the risk for liver injury. Despite the study design limitations, as well as the use of data sources such as spontaneous reporting databases, findings from our systematic review can be useful for the detection of rare adverse events, which has been recognized as a primary tool for pharmacovigilance reflecting the reality of clinical practice [19,50]. In addition, research findings on drug safety such as ours should spur on further experimental studies aimed at investigating the underlying mechanism and degree of severity of nimesulide-induced hepatotoxicity.
This systematic review has important implications for clinical practice. Currently, nimesulide is still available on the market in many countries (e.g., Bulgaria, Czech Republic, Greece, Hungary, Italy, Poland, Portugal, Romania, Slovakia and South Korea) despite its market withdrawal in several countries. Clinicians should consider prescribing nimesulide only as a second-line medication for the treatment of acute pain or dysmenorrhea and should monitor those patients with an underlying risk for liver injury from very early phase, even with shortterm use of nimesulide. Furthermore, an appropriate decision support system or vigilance teamwork including pharmacists would enable clinicians to better monitor nimesulide use and its associated adverse effects, especially in patients who concurrently use other potentially hepatotoxic drugs.

Conclusions
Our study indicates that nimesulide use is associated with an approximately twofold increased risk for hepatotoxicity. The association between nimesulide use and related hepatotoxicity is supported by our comprehensive disproportionality analysis, showing an increased rate of reported hepatic adverse events with nimesulide, compared with other NSAIDs. Further studies of nimesulide-induced hepatotoxicity are needed to evaluate the risk, as well as to better quantify the absolute risk, for hepatotoxicity associated with nimesulide by age, gender, and treatment dose and duration.