Acetylsalicylic acid and other salicylates in relation to Stevens–Johnson syndrome and toxic epidermal necrolysis

Abstract

Aims

Various nonsteroidal anti-inflammatory drugs are known to increase the risk of Stevens–Johnson syndrome and toxic epidermal necrolysis. The relationship between salicylate treatment and these conditions is not known.

Methods

A case-control study was conducted in four countries in Europe from 1989 to 1995.

Results

Among 373 cases and 1720 controls, the multivariate relative risk estimate for any salicylate use in the previous week was 1.3 (95% confidence interval, 0.8–2.2); no statistically significant elevations were observed for single ingredient preparations or for salicylate-containing combination products.

Conclusions

Acetylsalicylic acid and other salicylates are not associated with a measurable increase in the risk of these rare but severe reactions.

Introduction

Stevens–Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are rare life-threatening cutaneous reactions that are known to be induced by various medications, including some nonsteroidal anti-inflammatory drugs (NSAIDs) [1]. Many drugs have been implicated based solely on case reports, without more definitive evidence from formal epidemiological studies. Although acetylsalicylic acid (ASA) and other salicylates have not previously been suspected, they are commonly used analgesics, and given the concerns about other NSAIDs, we have evaluated the relationship between their use and the occurrence of SJS and TEN, using data from an international case-control study conducted in France, Germany, Italy, and Portugal from 1989 to 1995 [2, 3].

Methods

Potential cases (individuals admitted to hospital with blistering erythematous lesions) were identified by contact between study personnel and burn units, dermatology wards, paediatric wards, and intensive care units. Information on the clinical course, skin biopsies, and photographs were reviewed by a panel of study dermatologists without knowledge of exposures. The certainty of the diagnosis was classified as possible, probable, or definite, according to an algorithm which weighted the sources of information. Cases were classified as SJS, overlap between SJS and TEN, or TEN, based on the type and distribution of lesions and the extent of involvement [4]. Controls were selected from among other hospitalized patients in the same regions, according to a list of diagnoses which included acute dramatic infections (e.g. peritonitis), trauma, and acute abdominal emergencies (e.g. renal calculi), and matched to cases for age and gender. If there were no patients available with the preferred diagnoses, persons admitted for elective treatment, such as cosmetic surgery or hernia repair, were selected. Information on drug exposure prior to admission, along with demographic data and medical history, was obtained from patient interviews which utilized structured questions and prompted recall by asking about a wide range of indications for drug use. In the analysis, an index day was defined as the reference point for medication use. For the cases it was the day on which the first symptom of the reaction occurred, and for the controls, it was the day of the first symptom of the condition which resulted in the hospitalization.

Included in the analysis were 373 cases classified as definite (278) or probable (95) SJS or TEN. The diagnosis was considered only to be possible for 34 subjects, and these were excluded, along with 181 enrolled subjects who were determined to have other diagnoses, and 100 classified as erythema multiforme. There were 1720 controls, including those matched to potential cases that were subsequently excluded. This maximized the number of subjects to provide as much information as possible on uncommonly used drugs. While the case and control series were still broadly comparable according to age, sex, and region, the exact matching was no longer applicable because of the extra controls. Potential confounding by concomitantly used drugs and other factors was therefore taken into account by unconditional logistic regression [5]. The following factors were included in the multivariate model: age, sex, region, reliability of the subject (as judged by the interviewer), history of radiotherapy, collagen vascular disease, herpes, HIV infection, use in the week before the index day of paracetamol, glafenine, pyrazolones, oxicam NSAIDs, propionic acid NSAIDs, diclofenac, other NSAIDs (as a group), chlormezanone, allopurinol, corticosteroids, anti-infective drugs, sulphonamides, anticonvulsants, any other drugs suspected of causing SJS or TEN, and any other drugs.

Results

A total of 47 cases and 133 controls had used salicylates in the week before the index day (Table 1). The overall crude relative risk estimate was 1.7, and the multivariate estimate was 1.3 (95% confidence interval, 0.8–2.2). For ASA taken in a single ingredient preparation, the multivariate estimate was 1.5 (0.9–2.7), and for combination products containing ASA, it was 1.6 (0.4–6.3). The estimate for single ingredient preparations of salicylates other than ASA was 0.7 (0.2–2.7). There were very few users of combination products containing non-ASA salicylates. When the data were divided according to the specific type of reaction, there were no significantly elevated relative risk estimates (Table 2). The point estimate for combination products containing ASA in relation to overlap SJS/TEN was 3.6, but it was based on small numbers, with a lower confidence limit of 0.6.

Table 1.

Use of ASA and other salicylates among 373 cases of SCAR and 1720 controls according to type of preparation.

Drug	Type of preparation	Cases	Controls	Crude relative risk estimate	Multivariate relative risk estimate (95% confidence interval)
Any salicylates	–	47	133	1.7	1.3 (0.8,2.2)
ASA	Single ingredient	31	92	1.6	1.5 (0.9,2.7)
	Combination	6	12	2.4	1.6 (0.4,6.3)
Other salicylates	Single ingredient	8	26	1.5	0.7 (0.2,2.7)
	Combination	2	3	–	–

Table 2.

Use of ASA and Other Salicylates Among 373 Cases of SCAR and 1720 Controls According to Type of Preparation and Disease Classification.

Class	Drug	Type of preparation	Cases	Controls	Crude relative risk estimate	Multivariate relative risk estimate (95% confidence interval)
SJS	ASA	Single ingredient	12	92	1.6	1.2 (0.5,2.6)
		Combination	2	12	–	–
	Other salicylates	Single ingredient	3	26	1.4	1.4 (0.3,4.6)*
		Combination	1	3	–	–
Overlap	ASA	Single ingredient	7	92	1.3	1.7 (0.6,4.6)
SJS/TEN		Combination	4	12	5.7	3.6 (0.6,21)
	Other salicylates	Single ingredient	3	26	2.0	2.0 (0.4,6.6)*
		Combination	1	3	–	–
TEN	ASA	Single ingredient	12	92	2.1	1.5 (0.6,3.9)
		Combination	0	12	–	–
	Other salicylates	Single ingredient	2	26	–	–
		Combination	0	3	–	–

^*Crude estimate.

Discussion

The present results, derived from a large, carefully conducted epidemiologic study, are sufficient to rule out large increases in risk of SJS and TEN following salicylate use. The multivariate relative risk estimates, for which confounding was controlled, were reasonably close to 1.0 with upper confidence limits that were sufficient to exclude a 2.3-fold increase in risk for salicylates overall, and 2.8-fold for single ingredient preparations containing ASA or other salicylates. Although the data were less stable for combination preparations and when the cases were divided according to type of reaction, there were no significantly elevated multivariate estimates. In addition to confounding being taken into account, potential biases that normally apply to studies of this type are more likely to lead to false positive associations than to real associations being missed. Of particular concern in a case-control study of a disease known to be induced by drugs is the possibility that controls may tend to underreport their exposures due to memory loss, whereas cases are likely to have their memories primed from prior questionning by medical personnel and thus give more complete histories of drug use. The present study was designed and conducted to minimize such differential reporting, and the results that were obtained are not plausibly explained by this type of bias.

In contrast to salicylates, other NSAIDs are known to increase the risk of SJS and TEN [1]. Perhaps the most striking example is provided by oxicams, which have been shown in data from the present study [3] to have a relative risk of 72 within the first 2 months of use. We conclude that both ASA and other salicylates, when taken either in single ingredient products or in combinations with other compounds, are not associated with a substantial increase in the risk of SJS or TEN, although the possibility of modest increases cannot be ruled out. Given the rarity of the conditions, any increase in relative risk that is not excluded could only lead to a very small absolute risk of developing SJS or TEN.