Stevens–Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are severe cutaneous disorders characterized by acute skin blisters and mucous membrane erosions. They are severity variants of drug reactions that result in necrosis of the epidermis and other epithelia. According to current definitions  the main difference between the two is the extent of skin detachment: < 10% for SJS and > 30% for TEN (Fig. 1).
Because SJS and TEN are very rare, the risk cannot be evaluated in cohorts of treated patients and case–control studies are considered to be more accurate. To assess the risk of SJS or TEN in the context of drug treatment we have been conducting, since May 1997, a case–control study in Austria, France, Germany, Holland, Italy and Israel (EuroSCAR study). Risks are evaluated by comparing the rates of exposure to drugs between case patients and the controls. We present here the data collected in HIV-infected patients enrolled in that study up to 1 November 1999.
Patients and methods
Cases of SJS and TEN were actively detected through regular contact with a network of relevant facilities at which cases were treated. All cases requiring hospital admission were included. The detection network covered a population of about 120 million.
Selection criteria were: hospitalization with a diagnosis of SJS or TEN; widespread exanthema with ≥ 1% detachment; more than one blister; not only acral extension; presence of mucous membrane erosions. Potential cases were interviewed by trained investigators to determine exposure to drugs in the 4 weeks before admission (or before the onset of the disease in cases occurring within a hospital).
Clinical data, clinical photographs (available in 90% of cases) and skin biopsies (available in 70% of cases) of potential cases were reviewed by experts blinded to drug exposure data and who: ascertained potential cases as definite, probable, possible or excluded; classified the cases as SJS (detachment < 10% of the body surface area), overlap SJS–TEN (detachment 10–29%) or TEN (detachment ≥ 30%); and determined the most probable date of onset (index date). Only patients with a definite or probable diagnosis were included in the study.
Three hospital controls matched for age and sex were obtained for each case among persons admitted to the same hospital with conditions which were not expected to result from drug use (e.g. trauma, pneumonia, appendicitis, hernia).
Standardized information was obtained from cases and controls concerning medications, recent infections, demographics and relevant medical history. Data collection on medications used a list of indications (e.g., pain, headache, cough) and a list of trade-names of the main ‘suspect’ drugs.
Clinical characteristics and drug exposures before the index date were checked in all patients known to be infected by HIV-1.
Because with standard case–control analysis the crude odds ratio (OR) was ‘infinite’ for nevirapine, we used exact logistic regression, median unbiased estimate calculated according to Hirji et al. , and lower limit of the confidence interval according to Thomas .
The control group extracted from the general population was not optimal for evaluating drugs only indicated for the treatment of HIV infection. To overcome this limitation we performed a case-crossover analysis. The principle of this analysis is to compare the rate of exposure to a risk factor at different periods of time before the occurrence of an acute event [4,5]. In this study exposures during a ‘case period’ of 7 days before the index date were compared with exposures in a ‘control period’ of 7 days, 3–4 weeks earlier. Therefore each case served as his own matched control. The OR was the ratio of cases exposed during the ‘risk period’ only to cases exposed during the ‘control period’ only. When the ratio was infinite an exact binomial confidence interval was calculated.
Between May 1997 and November 1999 364 potential cases and 874 potential controls were enrolled. A probable or definite diagnosis of SJS or TEN was established in 246 cases. Among them 18 were known to be infected by HIV-1 (7.3%): three with SJS, seven with SJS/TEN overlap and eight with TEN. Two of 18 died as a result of the reaction. Among the 246 SJS and TEN patients 57 died (23%). Because the study was limited to the acute phase of the disease, information on sequelae was not available.
Recent or recurrent herpes was not associated to the reactions (0/18 and 2/18 respectively versus 6% and 24% of controls.)
Exposure to nevirapine was present in 15 out of 18 patients (83%). The other three patients were respectively exposed to: diclofenac alone; fluconazole and mintezol; amoxicilline + clavulanic acid, allopurinol and efavirenz, all introduced recently, plus clonazepam, cotrimoxazole, stavudine and abacavir for longer than 2 months.
The characteristics of the 15 patients exposed to nevirapine are presented in Table 1. They were four women and 11 men aged 21–59 years (median, 35 yeras), 10 were from France, three were from Germany, one was from the Netherlands and one was from Italy. The most recent counts of CD4 cells ranged from 4 to 1033 × 106/l (median, 234 × 106/l). All patients had mucous membrane erosions. The detachment of epidermis involved 4–55% of the body surface area (median, 25%). One patient died from SJS/TEN overlap.
All but one of the 15 patients exposed to nevirapine received other antiretroviral agents, introduced on the same date as nevirapine in 12 cases. No specific drug combination appeared among these associated drugs (stavudine, five; didanosine, five; zidovudine, four; lamivudine, four; abacavir, two; ritonavir, one; nelfinavir, one).
The reaction began 10–240 days after the introduction of nevirapine (median 12 days). All patients had initially received a daily dose of 200 mg (one tablet) according to the recommendation of a lead-in period. For 10 out of 15 patients the reaction began when they were still taking this initial dosage.
As no control among the 676 validated was exposed to nevirapine the OR was infinite, with a median unbiased estimate of 62 and a lower value of the 95% confidence interval of 10.4.
Details of the case-crossover analysis are shown in Table 2. Nevirapine was the only antiretroviral agent significantly associated with an increased risk of SJS or TEN.
We were not able to analyse the effect of treatments (corsticosteroids or high dose human immunoglobulins) given for SJS or TEN because this information was obtained for only a minority of patients.
HIV infection increases the risk of drug eruptions  including SJS and TEN [7,8]. In a previous case–control study of SJS and TEN in Europe, 7% of patients were HIV infected . In HIV infected patients most cases were related to antibacterial sulphonamides [6,9] in Western countries and to thiacetazone  in Africa.
The introduction of highly active antiretroviral therapy (HAART) resulted in a dramatic reduction of opportunistic infections. The lower exposure to antibacterial sulphonamides did not lead to the expected decrease of SJS or TEN in HIV-infected persons. Among cases of SJS and TEN enrolled in this study during the last 3 years, the percentage of patients known to be HIV infected was exactly the same as it was 10–5 years ago  (7.3% and 7% respectively). Our data strongly suggest that this persistence of a high risk of SJS or TEN in relation to HIV infection is currently associated with exposure to nevirapine.
The reasons why HIV-infected patients are at increased risk of severe cutaneous reactions are unclear. The mechanisms probably involve drug-specific cytotoxic lymphocytes. Therefore one may suspect that it could be part of an ‘immune restoration syndrome'. We consider this hypothesis to be unlikely as in previous years many cases were related to sulfadiazine treatment of central nervous system toxoplasmosis in profoundly immunosuppressed patients. In addition, if these reactions were related to immune restoration they should be observed with a variety of HAART regimens and not be restricted to those including nevirapine.
A high risk of severe cutaneous reactions had been detected in pre-marketing trials of nevirapine, but the benefits of this new antiretroviral agent were considered to outweigh the risk of SJS or TEN. In addition, it was suggested that the observance of a lead-in period decreased the risk of skin reactions [11,12]. However, in two-thirds of the cases exposed to nevirapine in the present study, the reaction occurred during the period of initial low dosage suggesting that escalating doses did not protect against SJS or TEN.
Whether the administration of systemic corticosteroids in addition to nevirapine during the lead-in period might decrease the rate of cutaneous reactions remains unknown [13,14].
Efavirenz and delavirdine, other non-nucleoside reverse transcriptase inhibitors (NNRTI), were released in Europe a few months after nevirapine . These drugs have not been used in as many people as nevirapine. In France at the end of 1999 about 15% of HAART regimens included nevirapine and 10% included efavirenz (D. Costagliola, French Hospital Database on HIV; personnal communication, January 2001). In our study we saw a single case (SJS/TEN overlap) exposed to efavirenz and none exposed to delavirdine. Our data strongly suggest that the risk of SJS or TEN is lower with efavirenz or delavirdine than with nevirapine.
We advise physicians to consider seriously the risk of these life-threatening cutaneous reactions when prescribing a HAART regimen. When nevirapine has advantages over another NNRTI, they must inform their patient of the risk of cutaneous reactions and provide clear guidelines of what to do in the case of skin eruption.
In HIV-infected patients a frequent practice is to ‘treat through’ mild eruptions and to withdraw the suspected drug only if markers of seriousness are present. Derived from accumulated clinical experience with sulphonamides in pneumocystosis, this attitude was justified by a lower effectiveness of alternative treatments and by the short elimination half-life of sulfamethoxazole. Concerning nevirapine a similar recommendation was issued by the European Agency for the Evaluation of Medicinal products: ‘nevirapine must be permanently discontinued in patients developing a serious cutaneous reaction’ (EMEA/11260/00, London 12 April 2000).
Withdrawing a drug with a long half-life when signs of a serious cutaneous reaction are already obvious may be too late. Actually the mortality of patients hospitalized for SJS or TEN was not improved if suspected drugs with long half-lives were withdrawn as soon as a diagnosis of SJS or TEN was made .
Considering the high risks of severe cutaneous adverse reactions associated with nevirapine; the long elimination half-life of this drug (25–30 h); and the existence of alternative drugs with lower risk of severe skin reactions, we suggest reconsideration of the ‘treating through’ attitude and recommend withdrawing nevirapine if any cutaneous eruption occurs during the first month of treatment.
In addition to the authors of the present paper the following members of the EuroSCAR group contributed to enrolment and evaluation of cases. Austria: A. Sidoroff; France: A. Flahault, A. Auquier, B. Sassolas, L. Vaillant, H. Bocquet, A. Dupuy; Germany: E. Schöpf, S. Anatkov, O. Hering, J. Schlingmann, B. Schneck, W. Schröder, E. Tas; Italy: L. Magrini, G. Manfredi; Israel: S. Halevy, A. Cohen; The Netherlands: C. Wensveen, V. Williams-Snijders.
1. Bastuji-Garin S, Rzany B, Stern RS, Shear NH, Naldi L, Roujeau JC. Clinical classification of cases of toxic epidermal necrolysis, Stevens–Johnson syndrome, and erythema multiforme. Arch Dermatol 1993, 129: 92–96.
2. Hirji KF, Tsiatis AA, Mehta CR. Median unbiased estimation for binary data. Am Stat 1989, 43: 7–11.
3. Thomas DG. Exact confidence limits for the odds ratio in 2 × 2 table. Appl Stat 1971, 20: 105–110.
4. Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol 1991, 133: 144–153.
5. Redelmeier DA, Tibshirani RJ. Association between cellular-telephone calls and motor vehicle collisions. N Engl J Med 1997, 336: 453–458.
6. Carr A, Tindall B, Penny R, Cooper DA. Patterns of multiple-drug hypersensitivities in HIV-infected patients. AIDS 1993, 7: 1532–1533.
7. Rzany B, Mockenhaupt M, Stocker U, Hamouda O, Schöpf E. Incidence of Stevens–Johnson syndrome and toxic epidermal necrolysis in patients with the acquired immunodeficiency syndrome in Germany. Arch Dermatol 1993, 129: 1059.1059.
8. Roujeau JC, Kelly JP, Naldi L. et al
. Medication use and the risk of Stevens–Johnson syndrome or toxic epidermal necrolysis. N Engl J Med 1995, 333: 1600–1607.
9. Carr A, Swanson C, Penny R, Cooper DA. Clinical and laboratory markers of hypersensitivity to trimethoprim-sulfamethoxazole in patients with Pneumocystis carinii pneumonia and AIDS
. J Infect Dis 1993, 167: 180–185.
10. Nunn P, Kibuga D, Gathua S. et al
. Cutaneous hypersensitivity reactions due to thiacetazone in HIV-1 seropositive patients treated for tuberculosis. Lancet 1991, 337: 627–630.
11. Pollard RB, Robinson P, Dransfield K. Safety profile of nevirapine, a nonnucleoside reverse transcriptase inhibitor for the treatment of human immunodeficiency virus infection. Clin Ther 1998, 20: 1071–1092.
12. Barner A, Myers M. Nevirapine and rashes. Lancet 1998, 351: 1133.1133.
13. Rey D, Partisani M, Krantz V. et al
. Prednisolone does not prevent the occurrence of nevirapine-induced rashes. AIDS 1999, 13: 2307.2307.
14. Barreiro P, Soriano V, Casas E. et al
. Prevention of nevirapine-associated exanthema using slow dose escalation and/or corticosteroids. AIDS 2000, 14: 2153–2157.
15. Garcia-Doval I, LeCleach L, Bocquet H, Otero XL, Roujeau JC. Toxic epidermal necrolysis and Stevens–Johnson syndrome: does early withdrawal of causative drugs decrease the risk of death? Arch Dermatol 2000, 136: 323–327.