Spontaneous reporting systems for adverse drug reactions (ADRs), such as the yellow card system run by the Medicines and Healthcare products Regulatory Agency (MHRA) in the UK and the US Medwatch system, are handicapped by under-reporting and provide limited detail on individual cases. In an effort to learn more about ADRs associated with disease modifying anti-rheumatic drugs (DMARDs) we set up a local reporting system seven years ago1 and reported preliminary findings in 2002 2.

Case reports of serious hepatotoxicity associated with sulfasalazine are common yet only 161 ADRs attributed to the liver and biliary system have been reported to the Medicines and Healthcare products Regulatory Agency (MHRA) including nine fatalities 3. In the US six patients developed drug related liver failure associated with sulfonamides in 2005 4. Drug 'hypersensitivity' reactions which include a morbilliform rash and hepatitis have, most recently, been called the DRESS (Drug Rash with Eosinophilia and Systemic Symptoms) syndrome 56 a term that places emphasis on eosinophilia. However serious toxicity without eosinophilia but including hepatitis is well recognised and older terms for this reaction, which occurs with other drugs such as carbamazepine and dapsone, have been called the 'three week syndrome' or the 'sulphone syndrome' 78.

Case reports are an important source of data which can guide management of new cases and trigger pharmaco-epidemiological investigations that may determine the probability of drug related toxicity. Drug related hepatotoxicity is an important cause of morbidity and mortality and indeed is the most common reason for withdrawing new drugs 9. In this investigation, which was triggered by the occurrence of liver failure in two of our patients taking sulfasalazine, we describe a series of patients who developed serious hepatotoxicity whilst taking sulfasalazine.

We have solicited reports of serious ADRs related to DMARDs from health care professionals treating patients with rheumatic disease in our institution by sending a postal or email postcard every 2 to 4 weeks over a period of 7 years. Details were requested for patients receiving DMARDs for RA or seronegative arthritidies. At inception we canvassed reports from colleagues in other local hospitals 2 but currently postcards and emails are targeted primarily at staff in our unit. We began in October 1999 when the first event was reported. A widely used definition of serious ADRs 10 was given in our early postcards but latterly any event regarded by a clinician as serious has been logged on our database. Clinicians are able to report serious events spontaneously or when prompted. This investigation was instigated when we became aware that two of our patients had developed liver failure whilst taking sulfasalazine.

Patients taking sulfasalazine and who developed a serious ADR which included hepatotoxicity were identified from our database. Identified cases were assessed for whether they met the definition of a severe or a life-threatening event according to the common toxicity criteria agreed by OMERACT (Outcome Measures in Rheumatology Clinical Trials) 11. Thus, patients had to have major functional impairment of any duration associated with a rise in transaminases of at least 3 times the upper limit of normal, or a similar rise in alkaline phosphatase, or both. Concomitant jaundice was not necessary. One senior clinician made these judgments (PJ). For those meeting these criteria a detailed case report was prepared. The pattern of hepatatoxicity according to serum liver enzyme changes was classified as hepatocellular (transaminases above 2× normal alone or ratio of transaminases to alkaline phosphatase ≥ 5), cholestatic (increase in alkaline phosphatase above 2× normal or when ratio of enzymes is ≤ 2) or mixed according to agreed international criteria 12. Case reports were reviewed by two senior clinicians who agreed, by consensus, whether the likelihood of drug-related hepatotoxicity was: highly probable; probable; possible; unlikely; or excluded. Two clinicians also used a detailed clinical index used for 'Causality Assessment of a Drug in a Case of Acute Liver Injury' 1213. This index was used in an effort to improve the probability that true drug related ADRs were identified 14. The index includes domains for temporal relationship between drug use and adverse event, course of illness, consideration of other risk factors and search for non-drug causes, concomitant drug use and effects on re-challenge, if done.

We estimated the frequency of serious hepatotoxicity associated with sulfasalazine by making a series of conservative estimates and by using published data on the incidence of rheumatic diseases and the likely use of sulfasalazine.

We estimated that the frequency of hepatitis with sulfasalazine was approximately 0.4% in our population. Two of our patients developed liver failure – an estimated rate of about 1.4 in 1000 people treated. It is possible that a higher than expected frequency of events occurred by chance during our surveillance and that the actual rate is lower: our estimates are in excess of those suggested by reviews of drug induced hepatotoxicity 22; rates reported in observational studies; and data reported to the Committee for Safety of Medicines in the UK. For example, Ransford and Langman reported that 8.6 hepatitis reactions occurred per million prescriptions in RA patients compared with 3.1 per million prescriptions for people with inflammatory bowel disease 23. Amos and colleagues, in another observational study, showed that only 2 of 774 patients stopped sulfasalazine because of abnormal liver enzymes 24. However 26% of patients ceased therapy in their study within 1 year because of toxicity and a majority because of nausea, abdominal pain and vomiting: symptoms that were commonly encountered in our series with hepatitis and which often led to uncertainties about diagnosis. Despite reservations about our estimates, under reporting of serious adverse events, even for drugs under special surveillance, is well known and only half of all cases identified by prescription event monitoring are notified spontaneously 25. Indeed a majority of the events related to sulfasalazine reported in this study (30 of 38) were from one institution indicating under-reporting from other centres involved at the beginning of our surveillance.

Most ingested sulfasalazine reaches the large bowel unchanged where it is split by colonic bacteria into sulphapyridine and 5-aminosalicylic acid (mesalazine). The latter remains in the large bowel but sulphapyridine is largely absorbed and eliminated in urine, principally after acetylation. Acetylation is determined by polymorphisms of the N acetyl transferase type II (NAT2) gene of which over 30 are known 26. Between 40 and 70% of European and African populations, and less than a third of Far Eastern populations, are slow acetylators. Toxicity due to sulfasalazine including hepatitis appears, in general, to be more common in slow acetylators 272829. Thus an excess of Black British individuals with an African or Caribbean background in our series cannot readily be explained by current data for NAT2 genotype and phenotype and other explanations need to be considered. We do not know whether the reported excess of hepatitis and leucopenia attributed to sulfasalazine in RA patients compared with that in people with inflammatory bowel disease 23 is due to disease, patient characteristics, or reporting bias because of differences in monitoring practices between rheumatologists and gastroenterologists.

Drug induced hepatotoxicity may occur as a direct result of a drug or it's metabolites on hepatocytes or via immune activation. Reactions to sulfasalazine are believed to arise because of an idiosyncratic delayed-type hypersensitivity reaction that may affect internal organs variably. Fatalities may occur in 10% or more cases and are commonly due to liver involvement 30. It is suggested that this is due to eosinophil infiltration of the liver 31; eosinophilia on liver biopsy was noted in one of our patients with liver failure. However it is also reported that there are no specific histological features associated with drug induced hepatitis 32. Treatment of drug related hepatotoxicity is supportive in most cases 30. Cessation of suspected drugs is vital and steroids are commonly used although there is no clear consensus that they are effective 31. Four of our patients were given steroids and the patient with liver failure who survived had received an intra-muscular injection of methylprednisolone just prior to starting sulfasalazine.

Toxicity without typical features of hypersensitivity, such as a skin rash, is well recognised 33. Adverse reactions to sulphonamides are more common in the presence of viral infections such as human immunodeficiency virus 34 and perhaps human herpes simplex-6 (HHV6) although the one patient we tested for HHV6 was negative 35. Systemic reactions to sulfasalazine and other drugs which include eosinophilia have been dubbed the DRESS syndrome. Associated interstitial nephritis is less common and likely to be due to mesalazine 23. In our series eosinophilia was not invariable and we concur with Peyriére and colleagues that this term is unnecessarily restrictive 36. The term 'three week syndrome' is older and accurately describes the timing of serious toxicity: it is worth noting that both the British Society for Rheumatology guidelines 37 and the summary of product characteristics for salazopyrin 38 recommend that liver function tests should be monitored every 4 weeks initially. Thus a majority of our patients would not have been identified by national monitoring recommendations. Our practice is prescribe sulfasalazine at 500 mg per day in week 1 and increase weekly by 500 mg up to 2 g per day at week 4. We monitor blood every 2 weeks for the first 12 weeks of therapy 39. This policy was in place at the time these events occurred and no deviations from our policy were identified. The occurrence of these events has led some local clinicians to argue for weekly monitoring during the first month of therapy especially in patients with a Black British background.

Determining causality, in cases of suspected drug induced hepatitis, can be difficult not least because of concomitant therapy including possible synergistic effects between potentially hepatotoxic drugs and alcohol use. One of our patients was known to consume alcohol to excess and a majority on other potentially hepatotoxic drugs such as diclofenac. Diagnostic work-up in our patients also varied greatly and, understandably, appeared to depend on the severity of liver disease. In some cases limited diagnostic testing appeared to reflect greater clinical certainty in the diagnosis of a drug induced reaction. However the absence of a comprehensive work up in all cases hampers retrospective judgments about causality. Various indices have been developed to assess causality which rely on a scoring system but there is no agreed gold standard and agreement between different scales is often poor 13. We tested the causality assessment index described by Danan and Benichou 12 (see Table 2), regarded as essential in drug induced hepatotoxicity studies 32. We also used a simple consensus method similar to the preferred judgments of hepatologists 40. There was a poor relationship between our consensus method and the causality assessment index, in part because of dependence on a positive re-challenge to attain maximum scores with the causality index. Only one of our patients was re-challenged and developed an adverse reaction with 48 hours. Clinicians should beware of re-challenging patients with suspected hypersensitivity reactions because of reports that even a small dose on rechallenge can precipitate massive hepatic necrosis 3241.

We conclude that serious hepatotoxicity often associated with a skin rash and variably associated with eosinophilia occurs within 6 weeks of starting sulfasalazine therapy. Our experiences indicate that such toxicity is likely to be missed by current national recommendations for monitoring. Toxicity in our series was particularly common in people of a Black British background of African or Caribbean descent. We believe that this adverse drug reaction is underappreciated and may occur with a frequency as high as 0.4%. Although this risk frequency is regarded by convention as uncommon the potential severity of reactions seen may make some patients and practitioners wary of sulfasalazine.

The author(s) declare that they have no competing interests.

PJ and RA collated the case reports, extracted data, assigned causality scores and drafted the abstract. PJ conceived the project and drafted this manuscript. FM compiled the database, sought adverse events and did initial analyses. DH, SB. ER, AF, KR, RJ and PJ contributed cases and participated in study discussions. All authors read and approved the final manuscript.