Is acetaminophen associated with a risk of Stevens–Johnson syndrome and toxic epidermal necrolysis? Analysis of the French Pharmacovigilance Database

Abstract

Aim

Stevens–Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are rare but severe mostly drug‐induced cutaneous reactions. Acetaminophen is an over‐the‐counter drug used worldwide to treat pain and reduce fever. In 2013, the US Food and Drug Administration informed the public that acetaminophen was associated with a rare risk of SJS/TEN. The aim of the present retrospective study was to analyse reports of acetaminophen as a possible suspect in the development of SJS/TEN from the French Pharmacovigilance Database (FPDB).

Methods

Cases of TEN/SJS with acetaminophen as a suspect drug registered in the FPDB, collected from January 2002 to December 2013, were analysed by an expert group. The algorithm of drug causality for epidermal necrolysis (ALDEN) was used as a reference tool for SJS/TEN to assess the causality of each suspect drug.

Results

After exclusion of 16 nonvalidated cases, 112 cases (47 TEN, 51 SJS, 14 SJS/TEN overlaps) involving 574 suspected drugs (5⋅1/case) were analysed. In 80 cases, the acetaminophen ALDEN score was inferior or equal to that of other drugs, associated with a higher suspicion for causality. In 32 cases, acetaminophen had the highest score but matched with a ‘very unlikely’ or ‘unlikely’ causality in 12 cases. For the 20 remaining cases with a ‘possible’ or ‘ probable’ causality, a protopathic or a confounding bias was likely in 14 cases.

Conclusions

After analysis of the French pharmacovigilance data using the ALDEN algorithm, we found no obvious SJS/TEN risk related to the use of acetaminophen in this large national series.

What is Already Known about this Subject

• Data extracted from the literature on acetaminophen causality evidence in toxic epidermal necrolysis (TEN) and Stevens–Johnson syndrome (SJS) remain debatable. In epidemiological studies as well as in case reports, the protopathic or confounding bias hypothesis is usually not considered and analysis leading to a suspicion of acetaminophen causality is not detailed.

What this Study Adds

• This rigorous individual analysis of a large pharmacovigilance series found no obvious SJS/TEN risk related to the use of acetaminophen.

• For adverse drug reactions, a specific reasoning based on a precise analysis of each case, including chronology and details about clinical symptoms and context, is essential to prove drug causality.

Introduction

Stevens–Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are severe cutaneous reactions of a single disease spectrum, both of which are characterized by extensive apoptosis‐related detachment of the epidermis and erosion of mucous membranes. Although SJS and TEN are rare (1⋅2–6 SJS cases per million of the population per year and 0⋅4–1⋅9 TEN cases per million of the population per year), these diseases constitute a real public health problem owing to a high mortality rate of 23% at 6 weeks and 34% at 1 year 1 or incapacitating long‐term sequelae 2.

The percentage of epidermal detachment with respect to body surface area distinguishes SJS (<10%) from TEN (>30%) and SJS/TEN overlap (10–30%), resulting in increasing severity and a worsening prognosis of the disease as the percentage of skin involvement increases. SJS/TEN can initially present as an influenza‐like syndrome with fever, ocular pain, and ear, nose and throat (ENT) pain that precedes the skin lesions 3, 4.

SJS and TEN are notorious adverse drug reactions. Common causes are allopurinol, anticonvulsants (carbamazepine, phenytoin, phenobarbital, lamotrigine), sulfonamide‐based drugs, sulfasalazine, nonsteroidal anti‐inflammatory drugs and nevirapine 5, 6. In order to improve the individual assessment of drug causality in SJS/TEN, a specific algorithm called the algorithm of drug causality for epidermal necrolysis (ALDEN) was constructed by the RegiSCAR network 7. However, in 20–30% of SJS/TEN cases, no suspect drug can be identified. Currently, only a small percentage of cases labelled ‘idiopathic’ cases can be attributed to infections, such as Mycoplasma pneumonia, especially in children.

Acetaminophen, available over the counter, is commonly used as an analgesic or antipyretic agent. In August 2013, the US Food and Drug Administration (FDA) informed the public that acetaminophen had been associated with a rare risk of SJS/TEN. A search of the FDA Adverse Event Reporting System (FAERS) database from 1969 to 2012 identified 91 cases of SJS/TEN and several cases were extracted from the medical literature, including two positive rechallenge cases 8, 9. Epidemiological studies showed discordant results 5, 6, 10, 11, 12. The aim of the present retrospective study was to analyse reports of acetaminophen as a suspected drug for SJS/TEN in the French Institutional Pharmacovigilance Database (FPDB), administrated by the Agence Nationale de Sécurité du Médicament et des Produits de Santé (ANSM).

Materials and methods

A search in the FPDB was carried out on 28 February 2014 to identify all the SJS/TEN cases that named acetaminophen (paracetamol) among the suspect drugs recorded in France between 1 January 2002 and 31 December 2013. Two preferred terms from the Medical Dictionary for Regulatory Activities (MedDRA) were selected: ‘toxic epidermal necrolysis’ and ‘Stevens–Johnson syndrome’.

Selected cases were validated by an expert group [dermatologists (B.L.V., M.J.J.P.) and senior pharmacovigilance staff (B.L.V., C.G., M.J.J.P., V.G., M.Z.). The criteria used for inclusion into the ALDEN protocol were: (i) a validated diagnosis of SJS/TEN through clinical symptoms, with compatible cutaneous histopathology and/or validation by a dermatologist. Despite the lack of a validated diagnostic score for SJS/TEN, cases qualified as ‘typical’ when both bullous cutaneous lesions and mucous erosions were present, and ‘atypical’ when only bullous cutaneous lesions or mucous erosions were mentioned; (ii) a reportable date for the onset of the reaction, defined as the ‘index day’, including a mandatory ‘probable index day’ (date of the onset of symptoms or signs that progressed within 3 days to definite erosions or blisters of the skin or mucous membranes) and an optional ‘earlier index day’ (onset of prodromes such as fever, influenza‐like syndrome, ENT symptoms); (iii) an accurate chronology of drug exposures. When necessary, additional information was requested from the regional pharmacovigilance centres concerned.

The ALDEN protocol was then applied by the expert group, working in pairs, to determine the final score for every suspect drug in each SJS/TEN case. Discrepancies were discussed within the whole group, to find a consensus. Drug notoriety was established for all suspect drugs according to the ALDEN method 7. To take the FDA alert into account, the notoriety value for acetaminophen was set at 1 (‘suspected drug, under surveillance’). The final score of ALDEN can range between −12 and 10. Drug causality related to the ALDEN score is defined as ‘very unlikely’ (score <0), ‘unlikely’ (score = 0–1), ‘possible’ (score = 2–3), ‘probable’ (score = 4–5) and ‘very probable’ (score ≥6).

Cases were then allocated into three groups regarding the acetaminophen ALDEN scores with respect to the other suspect drugs: group 1: ALDEN score for acetaminophen lower than that for all other suspect drugs; group 2: ALDEN score for acetaminophen equal to that for at least one other suspect drug; group 3: ALDEN score for acetaminophen greater than that for all other suspect drugs, or acetaminophen being the only suspect drug. A focus on group 3 cases was then made, to evaluate acetaminophen causality according to the ALDEN score and other possible aetiologies.

Results

The initial search in the FPDB identified 128 cases. After exclusion of 16 nonvalidated or unusable cases (Figure 1), 112 cases were analysed, including 47 TEN, 51 SJS and 14 SJS/TEN overlaps. Fifty‐four per cent of the patients were female and 20⋅5% were less than 18 years of age. The case characteristics are presented in Table 1. ALDEN scores were calculated for the 574 drugs (5⋅1/case) identified as suspect drugs in the 112 selected cases.

Figure 1.

SJS/TEN case selection and analysis. ALDEN, algorithm of drug causality for epidermal necrolysis; FPDB, French Pharmacovigilance Database; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis

Table 1.

Characteristics of cases

	SJS		SJS/TEN overlap		TEN		Total
	51		14		47		112
	Typical	Atypical	Typical	Atypical	Typical	Atypical	Typical	Atypical
	41	10	12	2	46	1	99	13
Male/female	27/24		7/7		17/30		51/61
Age
<18 years	11	0	1	0	11	0	23	0
≥18 years	30	10	11	2	35	1	76	13
Age (years)
Mean ± SD	36.9 ± 21		43.9 ± 18.9		43.1 ± 27.9		40.5 ± 24.4
Median (range)	33 (6–77)		37 (25–83)		44 (1–90)		37 (1–90)
IQR: 1–3	19–56		30–46		18.5–67.5		19.8–61.3
Death	1		1		14		16
Allocation according to ALDEN score for acetaminophen vs. other suspect drugs
Group 1	27	7	7	2	28	1	62	10
Group 2	3	1	2	0	2	0	7	1
Group 3	11	2	3	0	16	0	30	2

ALDEN, algorithm of drug causality for epidermal necrolysis; IQR, interquartile range; SD, standard deviation; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis

Group 1: ALDEN score for acetaminophen less than that for all other suspect drugs

Group 2: ALDEN score for acetaminophen equal to that for at least one other suspect drug

Group 3: ALDEN score for acetaminophen greater than that for all other suspect drugs, or acetaminophen being the only suspect drug

Analysis revealed that acetaminophen was not the most suspected drug in 72 cases (64%; group 1) and that it was a possible culprit along with another drug in eight cases (7%; group 2). The main suspected drugs with an ALDEN score ≥2 (causality at least ‘possible’) identified in group 1 and group 2 are listed in the Table 2. Analysis of the 32 cases belonging to group 3 (Figure 2) shows that acetaminophen had the highest ALDEN score among suspect drugs but matched with a ‘very unlikely’ or ‘unlikely’ causality in 12 cases. For the 20 remaining cases, acetaminophen causality was ‘possible’ in 14 cases and ‘probable’ in six cases. A protopathic or a confounding bias was highly probable in 14 cases owing to an acetaminophen intake either for prodromes of SJS/TEN or for onset of the causal infection of SJS/TEN (fever, influenza‐like syndrome). In these cases, the acetaminophen score was recalculated using an ‘earlier index day’, giving a ALDEN score of −5 matching with a ‘very unlikely’ causality.

Table 2.

Drugs with an ALDEN score ≥2 (causality at least possible) identified as more suspected than acetaminophen (group 1) or associated with an equal degree of suspicion for causality as acetaminophen (group 2)

Drugs	Group 1	Group 2
Penicillins	11	–
Amoxicillin	10
Piperacillin	1
Cephalosporins	10	–
Ceftriaxone	6
Cefixime	2
Cefpodoxime	1
Ceftazidime	1
Tetrazepam	7	4
Anticonvulsant drugs	6	–
Carbamazepine	3
Phenytoin	2
Phenobarbital	1
Sulfonamides	6	–
Cotrimoxazole	4
Sulfadoxine	1
Sulfasalazine	1
Oxicam NSAIDs	6	–
Piroxicam	5
Tenoxicam	1
Proton pump inhibitors	4	–
Omeprazole	1
Esomeprazole	1
Lansoprazole	1
Pantoprazole	1
Fluoroquinolones	4	–
Ciprofloxacin	3
Norfloxacin	1
Acetic acid derivative NSAIDs	4	–
Diclofenac	3
Etodolac	1
Glycopeptide antibiotics	2	–
Teicoplanin	2
Cycline antibiotics	2	–
Tetracycline	1
Doxycycline	1
Coxib NSAIDs	–	2
Celecoxib	–	2

ALDEN, algorithm of drug causality for epidermal necrolysis; NSAIDs, nonsteroidal anti‐inflammatory drugs

Figure 2.

Group 3 analysis. ALDEN, algorithm of drug causality for epidermal necrolysis; SJS, Stevens–Johnson syndrome

An M. pneumonia infection was proven in two paediatric cases (girls of 11 and 12 years of age, respectively) and a graft vs. host disease was possible in another case. In two other cases, involving a 67‐year‐old man with viral hepatitis C being treated with interferon and ribavirin, and an 81‐year‐old man receiving rituximab for lymphoblastic leukaemia, acetaminophen causality was debatable because of sporadic acetaminophen intake. In these latter two situations, but probably not uncommonly, prior acetaminophen intake without adverse effects was likely but not mentioned, resulting in a decrease in the ALDEN causality score. In the last case, SJS occurred 12 days after nimesulide, acetaminophen and dextropropoxyphene were administered, without protopathic bias or a confounding factor. Acetaminophen was the drug associated with a higher suspicion for causality because of the notoriety value, which was set at 1 for acetaminophen and at 0 for the two other drugs.

Discussion

The association between SJS/TEN and acetaminophen is widely debated in the literature. It is an intricate topic, accounting for: (i) the worldwide use of this over‐the‐counter antipyretic and analgesic drug; (ii) the low incidence of SJS/TEN; and (iii) possible biases such as protopathic bias (initiation of acetaminophen treatment occurs in response to fever or ENT pain, which are often the prodromal symptoms of SJS/TEN) or a confounding bias (initiation of acetaminophen treatment occurs in response to fever or ENT pain, which are related to an infectious disease accountable for SJS/TEN).

In the Serious Cutaneous Adverse Reaction (SCAR) case–control study, acetaminophen was found not to be a significant risk factor for SJS/TEN in France, whereas the multivariate risk was 9.3 in other countries (Germany, Italy and Portugal) 5. Differences in acetaminophen use among these countries may explain this discrepancy. At the time of this study, acetaminophen consumption in France was higher than in these other countries, and it was mainly used as an analgesic there, as opposed to as an antipyretic in the other countries. However, analysis restricted to subjects with fever did not explain the difference between these countries. It has been suggested that the high prevalence of use and the repeated exposure to acetaminophen in France could lead to either the selection of patients who do not react or the induction of tolerance. The EuroSCAR case–control study confirmed the huge disparity in acetaminophen exposure rates in different countries 6. Analyses showed a weak association between the risk for SJS/TEN and use of acetaminophen, with a multivariate relative risk of 1⋅9 (1⋅2–2.8), which did not vary significantly between countries, probably because of the lack of power to detect a difference. In order to limit the protopathic or confounding bias, the EuroSCAR authors restricted the analysis to patients who took acetaminophen at least 4 days before the index day, resulting in the removal of the previously observed association and casting the causal relationship into doubt. Similar results were seen for other drugs, such as pyrazolone analgesics, tramadol and nimesulide. In a Taiwanese case–control study using the national health insurance database and including bipolar inpatients, results regarding the association between acetaminophen and SJS/TEN were variable. What was observed, however, was an increased risk due to the combination of carbamazepine and acetaminophen. The underlying mechanism of such an interaction remains unclear 10. In the pooled analysis by using data of case subjects aged <15 years from the SCAR and EuroSCAR studies, a significant association between acetaminophen and SJS/TEN was highlighted in multivariate analysis, with an adjusted odds ratio of 5.0 (2.0–13.0), which remained significant after exclusion of cases involving exposure to acetaminophen at least 2 days before the index day to control a protopathic bias 11. In an Italian paediatric case–control study, a significantly elevated risk of SJS/TEN was found for acetaminophen, with an adjusted odds ratio of 3.2 (1.5–6.9), even after exclusion of cases in which the drug was started on the index day or in the 2 previous days 12. However, in these last two studies 11, 12, a confounding bias remains possible, especially as an infectious aetiology of SJS/TEN is probably more frequent in the paediatric population.

We found seven retrospective studies reporting an SJS/TEN series, mainly from Asian countries, in the literature 13, 14, 15, 16, 17, 18, 19. Acetaminophen was regarded as the suspect drug in 6–16% of cases but the assessment of drug causality was not detailed. It is thus impossible to exclude a protopathic or a confounding bias. In a study reporting 6364 SJS and TEN cases using the Japanese Adverse Drug Event Report database, 495 were attributed to acetaminophen but a protopathic or a confounding bias was likely, due to a far shorter time‐to‐onset period (<4 days) compared with the other, ‘classical’ SJS/TEN‐inducer drugs, such as allopurinol, lamotrigine or carbamazepine 13.

Twenty case reports of SJS/TEN related to acetaminophen are available in the international literature 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, two of them with a positive rechallenge. We were able to apply ALDEN retrospectively using available data in 12 cases. Acetaminophen causality was ‘probable’ (ALDEN score = 4) in four cases 21, 24, 29, 30, ‘possible’ (ALDEN score = 2–3) in five cases 20, 26, 30, 31, 32 and ‘very unlikely’ (ALDEN score <0) in three cases 23, 25, 33. The ALDEN score was inapplicable in eight cases because of a lack of data on the delay between acetaminophen intake and index day 22, 27, 28. The two positive rechallenges were questionable because of a dubious SJS diagnosis in the first case 20 and a conflicting immediate anaphylactoid reaction in the second case 21. Apart from the drug causality, a protopathic or confounding bias was possible in 15 cases (75%) due to an infectious context 21, 22, 23, 24, 25, 29, 30, 31, 33, including six cases with acetaminophen causality as ‘possible’ or ‘probable’. Moreover, in the twenty case reports, the diagnosis of SJS/TEN was often not clearly documented, and was even questionable, with clinical features often compatible with erythema multiforme major 20, which is more frequent in children and young adults, recurs regularly, has a better prognosis and can be either infectious or idiopathic in origin 3, 34, 35.

In accordance with the FDA Drug Safety communication published in August 2013, of the 91 cases of acetaminophen‐associated SJS/TEN identified in the FAERS in the space of 43 years, only six were categorized as probable, confirmed by a dermatologist and/or histological findings, without confounding medications administered within the 2 weeks preceding the events. The 85 other cases were categorized as possible cases. The delay between the start of acetaminophen treatment and the index day ranged from less than 24 h to 8 days 8. ALDEN scores were not specified for these reports. Indications for acetaminophen use varied between pyrexia and pain but the degree of fever or pain in each situation was not specified.

The search in the FPDB identified 112 SJS/TEN cases involving acetaminophen as a suspect drug in the space of 12 years. One of the main conclusions of the present study is that the application of ALDEN with regard to pharmacovigilance reports is feasible and of interest. With this in mind, reports of suspected SJS and TEN cases to the pharmacovigilance system should be sufficiently informative. It is essential to select validated SJS/TEN cases in order to apply the ALDEN protocol. Necessary information includes accurate data on the chronology of clinical events (particularly the index day) and of exposure to the drugs concerned. The precision of the FPDB data was good enough to select 87⋅5% of the SJS/TEN cases recorded by the national pharmacovigilance network. Case‐by‐case analysis of 112 cases using ALDEN resulted in 32 cases in which acetaminophen was the more likely or the only suspect drug. However, even in these cases, acetaminophen causality was eventually found to be questionable due to either another cause or a protopathic or confounding bias. The M. pneumoniae‐related cases were not excluded from the analysis because nondrug‐related possible causes of SJS/TEN are not among the ALDEN parameters.

The main discussion about acetaminophen causality in SJS/TEN, in this paper as well as in the litterature. In many cases, acetaminophen intake began a few days before SJS/TEN, even on the ‘probable index day’. This is why it may be important to take into account an ‘earlier index day’ rather than the index day defined by the onset of a obvious cutaneous or mucous injury. Indeed, according to ALDEN, a ‘suggestive’ delay from initial drug intake to index day can range from 5 days to 28 days. This delay is considered as ‘likely’ when a clinical event occurs 1–4 days after drug exposure and as ‘excluded’ when the drug exposure occurs on the index day.

The application of ALDEN for acetaminophen is difficult due to its status as an easily available over‐the‐counter drug, implying possible frequent self‐medication and widespread use both in children and adults, and resulting in less‐than‐accurate chronological and dosage analysis, particularly for a prechallenge evaluation.

Ultimately, despite a possible overestimated notoriety value of 1, only one SJS case with a ‘probable’ acetaminophen causality was identified because of a lower notoriety value for nimesulide and dextropropoxyphene.

In conclusion, our analysis of the French pharmacovigilance data using the ALDEN algorithm found no obvious SJS/TEN risk related to acetaminophen use in France. As such, the FDA alert was not confirmed in this large French series.

Competing Interests

There are no competing interests to declare.

The authors would like to thank Laurence Valeyrie‐Allanore for her expert assessment in some difficult cases, Cynthia Haddad for her help in learning and applying ALDEN and Pascal Auriche from ANSM for extraction of cases from the FPDB.

 

Lebrun‐Vignes, B. , Guy, C. , Jean‐Pastor, M.‐J. , Gras‐Champel, V. , Zenut, M. , and The French Network of Regional Centres of Pharmacovigilance and the French Investigators for Adverse Skin Reactions to Drugs (2018) Is acetaminophen associated with a risk of Stevens–Johnson syndrome and toxic epidermal necrolysis? Analysis of the French Pharmacovigilance Database. Br J Clin Pharmacol, 84: 331–338. doi: 10.1111/bcp.13445.