Toxic epidermal necrolysis: Five years of treatmentexperience from a burn unit

Bahar F. Firoz, MD, MPH,a Jeffrey Scott Henning, DO, MBA,c Lee Ann Zarzabal, MS,b

and Brad H. Pollock, MPh, PhDb

San Antonio, Texas

From

Ep

Sc

to

Fund

Conf

Acce

630

Background: Toxic epidermal necrolysis (TEN) is a serious drug eruption that results in death inapproximately 25% to 50% of patients. There is controversy over whether SCORTEN accurately predictsmortality or if treatment interventions such as intravenous immunoglobulin (IVIg) can alter mortality.

Objectives: We sought to determine whether SCORTEN accurately predicts mortality in this cohort,whether IVIg improved survival, and which drugs and medical comorbidities impacted mortality.

Methods: We summarize our experience prospectively over 5 years and 82 patients. Patients eitherreceived supportive care, intravenous immunoglobulin, or cyclosporine as treatment. All patients had aSCORTEN on admission, an offending drug on record, and a list of medical comorbidities.

Results: Of the 82 patients, 29% died from TEN. SCORTEN accurately predicted mortality in this cohortwith an area under the curve (AUC) of 0.83 in a receiver operator curve (ROC) analysis. A Kaplan-Meiercurve did not show improved mortality if patients received IVIg versus supportive care (P = .9).Medications most often responsible for TEN were trimethoprim/sulfamethoxazole, followed by anticon-vulsants, nonsteroidal anti-inflammatories, and allopurinol.

Limitations: This prospective cohort study design is not as ideal as patients presenting for a randomizedcontrolled trial.

Conclusions: SCORTEN was an accurate predictor of mortality in this cohort. Age older than 40 years,the presence of metabolic syndrome and/or gout, higher body surface area involvement, higherSCORTEN, and higher number of medical comorbidities statistically significantly increased risk of death.IVIg did not significantly alter mortality. Although the highest number of cases was due to trimethoprim/sulfamethoxazole, the greatest proportion of deaths was due to allopurinol. ( J Am Acad Dermatol2012;67:630-5.)

INTRODUCTION

Abbreviations used:

AUC: area under the curveBSA: body surface areaIVIg: intravenous immunoglobulinNSAIDs: nonsteroidal anti-inflammatory drugsOR: odds ratioROC: receiver operating curveTEN: toxic epidermal necrolysis

Toxic epidermal necrolysis (TEN), or Lyell’s syn-drome, was first described in 1956 as a scaldingeruption of the skin.1 TEN is a life-threatening drugeruption characterized by keratinocytic cell death atthe dermoepidermal junction in the skin andmucousmembranes, resulting in the scalded appearance ofthe skin.2 The eruption usually occurs 1 to 3 weeksafter initiation of the drug, and concomitant systemic

the Division of Dermatologya and the Department of

idemiology and Biostatistics,b University of Texas Health

ience Center, San Antonio; and the Department of Derma-

logy, Brooke Army Medical Center,c San Antonio.

ing sources: None.

licts of interest: None declared.

pted for publication December 9, 2011.

Reprint requests: Bahar Firoz, MD, MPH, Department of

Medicine/Division of Dermatology MC 7876, 7979 Wurzbach,

Third Floor Grossman, CTRC, San Antonio, TX 78229. E-mail:

firoz@uthscsa.edu.

Published online January 30, 2012.

0190-9622/$36.00

� 2012 by the American Academy of Dermatology, Inc.

doi:10.1016/j.jaad.2011.12.014

J AM ACAD DERMATOL

VOLUME 67, NUMBER 4Firoz et al 631

symptoms including fever, malaise, and skin pain arecommon. Although the incidence of TEN is very low,estimated between 0.4 and 1.2 per million personsannually, the mortality rate associated with this drugeruption is very high, estimated to be between 25%and 50% (average of 30%).3,4 SCORTEN is a scoringsystem developed to predict mortality in these

CAPSULE SUMMARY

d Toxic epidermal necrolysis (TEN) is adevastating drug eruption with highmortality; there are conflicting reportsabout whether interventions canimprove mortality, and whetherSCORTEN accurately predicts mortality inthis group

d This study shows that use of IVIg(intravenous immunoglobulin) did notsignificantly improve survival. Risk ofdeath was different depending on thedrug responsible for TEN. Age over 40years, the presence of metabolicsyndrome and/or gout, higher bodysurface area involvement, higherSCORTEN, and higher number of medicalcomorbidities were all statisticallysignificantly associated with a higher riskof death

d SCORTEN can be used to predictmortality in TEN patients. Although IVIgmay not improve survival in this cohort,new parameters such as offending drug,or medical comorbidities may be used tocharacterize patients at highest risk fordeath.

patients and includes param-eters such as age, body sur-face area, and certain bloodchemistries.5,6

Although the pathogene-sis of TEN is only partiallyunderstood, it is believedthat the epidermal necrosisis due to massive keratino-cyte apoptosis.7 This isthought to occur in largepart because of the interac-tion of the Fas- andFas-ligand pair, which areexpressed in epidermal ke-ratinocytes.8 Interventionsaimed at inhibiting this path-way have attempted to re-duce mortality of thisdisease, unfortunately withmixed results. Intravenousimmunoglobulin (IVIg) andcyclosporine both have po-tential in mitigating this dis-ease and improving survivalwhen compared with sup-portive care alone.9-14

TheBrooke-ArmyMedicalCenter (BAMC) in SanAntonio, Texas is a tertiarycare hospital with the onlyburn center in South Texas.

All local TEN cases are transferred to the burn unitfor treatment, where TEN-specific protocols areemployed. The burn unit uses a multidisciplinaryapproach involving surgery, dermatology, ophthal-mology, and gynecology or urology. We sought tosummarize our prospective experience over 5 yearswith 82 consecutively accrued patients. Patientseither received supportive care, IVIg, or cyclosporineas treatment protocols. All patients received anassessment of SCORTEN, prior exposure to offendingdrugs, and prevalent comorbidities.

METHODSBAMC Burn Unit TEN protocol

All patients admitted to our burn unit had blisterswith a confirmatory biopsy. The standard TEN

protocol in the BAMC Burn Unit is to terminate theuse of the offending drug immediately. No antibi-otics are prescribed unless the patient has confirmedpositive blood cultures. Fluid and electrolyte man-agement consists of aggressive monitoring and targeturine output is maintained between 0.5 and 1 mL perkilogram per hour (for a person weighing 70 kg, 40

cc/hr); over-resuscitation isavoided. Enteral nutrition isstarted as soon as possibleand titrated on the basis ofthe total body surface areaaffected. Corticosteroids arediscontinued or tapered ifstarted before the patient’sadmission to the Burn Unit.Wound care is essential to agood outcome and the goalis to protect the skin. Whenpatients are admitted, theyare wrapped in Kerlix gauzeto approximately 1 inch inthickness, across the entirebody and head; a window isleft for the eyes, mouth, andgenitals. The Kerlix gauze issoaked with 0.5% silver ni-trate solution (AgNO3) twicedaily. Silver nitrate solution isused to promote healing andact as an antimicrobial agent.Since silver nitrate perma-nently stains surfaces black,the floors are protected withthick plastic and the hospitalbeds are reserved for TENpatients only. This dressingis changed every 48 hours,and during dressing changes

sloughed and/or dead skin is removed to prevent anidus of infection. If patients were admitted to ourBurn Unit within 72 hours of blistering, they receivedIVIg treatment (4 g/kg divided over 3 days) unlesscontraindicated. If they were admitted to our BurnUnit more than 3 days after developing the firstblister, they received supportive care. Cyclosporinehas been used in some cases with low body surfacearea involvement, absence of contraindications tocyclosporine, and a SCORTEN of 0 to 1.

Data accessionedAll patients presenting to the Burn Unit with

biopsy-proven TEN were collected prospectivelyfor this institutional review boardeapproved study.A total of 82 patients were enrolled. SCORTEN was

Table I. Demographic variables for patients in thisstudy

Demographic criteria

N = 82

Total, no. (%)

Male 44 (100)Female 38 (100)Alive 58 (70.7)Dead 24 (29.3)Metabolic syndrome 19 (23.2)No metabolic syndrome 63 (76.8)IVIg 23 (28)Cyclosporine 8 (9.8)Supportive 51 (62.2)Age* 45.1 (20.4)Days in Burn Unit* 15.2 (16.0)Body surface area involved* (%) 34.8 (26.1)

IVIg, Intravenous immunoglobulin.

*Data expressed as mean (standard deviation).

Table II. SCORTEN on admission

No. of

patients

Predicted

mortality

Actual

mortality

%

No. of

deaths %

No. of

deaths

SCORTEN, no. (%)0 5 1.2 0.06 0 01 14 3.9 0.55 0 02 25 12.2 3.05 16.0 43 20 32.4 6.48 40.0 84 17 62.2 10.57 64.7 115 1 85 0.85 100.0 1

Total 82 26.29 21.56 29.26 24

J AM ACAD DERMATOL

OCTOBER 2012632 Firoz et al

calculated on admission and at 48 hours for eachpatient. Demographic and clinical information wasrecorded for each patient, including age, sex, andmedical comorbidities. Medical comorbidities in-cluded cancer, seizure disorder, gout, HIV infection,other infection, metabolic syndrome, psychiatricdisorders, connective tissue disease, and arthritis.Treatment including supportive care alone, IVIg, orcyclosporine, and the offending drug was alsorecorded.

Statistical considerationsSimple descriptive statistics were used to

characterize demographic and clinical factors.Continuously distributed variables are described bythe mean plus or minus standard deviation (SD) andcategorical variables by number and percentage, n(%). Association with mortality was compared withboth medications as well as medical conditions usingFisher’s exact test. Unconditional logistic regressionanalysis was used to assess prognostic factors usedfor prediction of the risk of death. The magnitude ofassociation between a covariate and the risk of deathfor the logistic regression analyses was summarizedas the odds ratio (OR) and 95% confidence interval(95% CI). Predictive accuracy of the logistic regres-sion models was assessed by receiver-operatingcharacteristic (ROC) curves and the area under thecurve (AUC). Kaplan-Meier curves were constructedto evaluate the actuarial risk of death relative totreatment regimen. The log-rank test was used tocompare treatment regimens relative to survival.

RESULTSA total of 82 patients (44 male, 38 female) were

enrolled in this study. As shown in Table I, 24patients (29.2%) died of TEN, which is comparableto the average TEN mortality rate published in theliterature. Nineteen patients (23%) had metabolicsyndrome, as defined by prevalent hypertension,diabetes, and hyperlipidemia. Of the 82 patients, 23(28%) received treatment with IVIg, 8 (9.8%) re-ceived cyclosporine, and 51 (62.2%) received stan-dard supportive care. The mean age plus or minusthe SD of the patients was 45 6 20 years and theaverage body surface area involved was 30%.Patients spent an average of 15 days in the BurnUnit. Of those who survived, the average stay waslonger at 17 days, compared with 11 days in patientswho died of TEN (P = .03).

SCORTEN was calculated for all patients onadmission, as presented in Table II. Of patientswith the lowest SCORTEN (score 0 and score 1), all19 survived. Twelve of the 18 patients (66%) withSCORTEN score of 4 or 5 died. In order to analyze

how accurately SCORTEN predicted mortality in ourcohort, we compared the probability of death foreach patient to a logistic model of the patient’s actualdeath status. The probability of death for eachpatient was calculated using the same logistic re-gression equation created by Bastuji-Garin et al.5

Fig 1 shows an ROC curve. An AUC of 1 indicates aperfect test, or perfect accuracy (100%), whereas anAUC of 0.5 means that the test is no better than arandom guess (50%-50%). In our cohort, the AUC forthe variable SCORTEN on admission was 0.83,showing excellent correspondence between ob-served and predicted mortality. We calculatedSCORTEN at 48 hours, which also had a highpredictive accuracy with an AUC of 0.80. Therewere no statistically significant differences (P = .31-1.0, exact binomial test of proportions) between thenumber of observed deaths and the number ofpredicted deaths at SCORTEN levels of 0 to 5,inclusive. Similarly, there was no statistically

Fig 1. ROC curve for SCORTEN at admission and 48hours. The probability of death was calculated for eachpatient by using the Bastuji-Garin model and comparedwith the actual death in this cohort by representation of anROC curve.

Table III. Offending medication and major medicalproblems shown in order of highest to lowest totalfrequency

Alive,

no. (%)

Dead,

no. (%)

Total,

no. (%)

Medication,* no. (%)Trimethoprim/sulfamethoxazole

23 (39.7) 7 (29.2) 30 (36.6)

Anticonvulsant 9 (15.5) 3 (12.5) 12 (14.6)NSAID 11 (19) 1 (4.2) 12 (14.6)Allopurinol 3 (5.2) 7 (29.2) 10 (12.2)Penicillin/cephalexin 7 (12.1) 1 (4.2) 8 (9.8)Antihypertensive 2 (3.4) 1 (4.2) 3 (3.7)Ciprofloxacin/levofloxacin 1 (1.7) 2 (8.3) 3 (3.7)Other (voriconazole,azithromycin)

2 (3.4) 0 (0) 2 (2.4)

Vancomycin/Lenazolid 0 (0) 1 (4.2) 1 (1.2)SSRI 0 (0) 1 (4.2) 1 (1.2)

Major medical problem,y

no. (%)Infectious 27 (46.6) 7 (29.2) 34 (41.5)Gout 4 (6.9) 8 (33.3) 12 (14.6)Seizure disorder 8 (13.8) 2 (8.3) 10 (12.2)HIV infection 5 (8.6) 2 (8.3) 7 (8.5)Arthritis 7 (12.1) 0 (0) 7 (8.5)Psychiatric 2 (3.4) 2 (8.3) 4 (4.9)Connective tissue disease 1 (1.7) 3 (12.5) 4 (4.9)Cancer 2 (3.4) 0 (0) 2 (2.4)Metabolic syndrome 2 (3.4) 0 (0) 2 (2.4)

NSAID, Nonsteroidal anti-inflammatory drug; SSRI, selective

serotonin reuptake inhibitor.

*P = .01; Fisher’s exact test.yP = .013; Fisher’s exact test.

J AM ACAD DERMATOL

VOLUME 67, NUMBER 4Firoz et al 633

significant difference in the overall predicted mor-tality of 22 patients (26%) and the overall observedmortality of 24 patients (29%) (P = .61, Fisher’sexact test).

The medications responsible for TEN are dis-played in Table III in order of highest to lowestoccurrence. The 4 highest offenders were trimetho-prim/sulfamethoxazole, anticonvulsants, nonsteroi-dal anti-inflammatory medications (NSAIDs), andallopurinol. Although 30 patients had TEN fromtrimethoprim/sulfamethoxazole, only 7 of these pa-tients died (23%). In comparison, 10 patients hadTEN from allopurinol, but 7 of these patients died(70%). Of the 12 patients with TEN due to NSAIDs,only 1 of these patients died (8.3%). There was astatistically significant association between the out-come of death and the offending medication (P =.01). None of the patients in this cohort developedTEN due to a pre-existing, documented drug allergy.However, as we are a tertiary referral center, it ispossible that preexisting drug allergies were notrelayed on admission to the Burn Unit. The timecourse to development of TEN also differed depend-ing on the offending drug and is as follows: trimeth-oprim/sulfamethoxazole, 5-14 days; NSAIDs, 7-20days; allopurinol, 5-28 days; penicillin, 6-12 days;angiotensin converting enzyme inhibitors, 10-31days; selective serotonin reuptake inhibitors, 10-25days; phenytoin, 11-41 days; lamotrigine, 12-38 days.

Table III also shows the predominant prevalentcomorbidities at the time of entry into the cohort.The top 3 medical problems included infectious

etiologies, gout, and seizure disorders. Of the 34patients with infectious reasons for TEN, only 7(20%) died, whereas of the 12 patients with gout, 7(67%) died of TEN. There was a significant associa-tion between outcome of death and the predominantprevalent medical conditions (P = .013).

The treatment options, which consisted of sup-portive care, IVIg, or cyclosporine, were comparedfor overall survival. The Kaplan-Meier survivalcurves for IVIg versus supportive care are shown inFig 2. There was no statistically significant differencein survival if patients received standard supportivecare versus IVIg therapy (P = .9, log-rank test). Whenall 3 treatment options were compared, there was nosignificant difference in survival (P = .15, log-ranktest).

Several different factors were evaluated for cu-mulative probability of death in a univariate logisticregression (Table IV). Patients older than 40 yearshad a higher risk of death (OR = 4.57, 95% CI = 1.58-13.22, P = .005). Patients with a higher number ofcomorbidities ([2) had more than a 9-fold increased

Fig 2. Kaplan-Meier survival curves by treatment regimen.The treatment options IVIg and supportive care werecompared for overall survival and are represented withKaplan-Meier curves. The treatments were comparedusing the log-rank test and were not significantly different(P = .9).

Table IV. Unadjusted logistic regression models forprediction of cumulative mortality

Odds ratio

(95% CI)

P

value

Admit age 1.04 (1.01-1.07) .003Age[40 yr (over 40 vs below) 4.57 (1.58-13.2) .005Treatment (IVIg vs supportive) 0.88 (0.30-2.53) .81Body surface area involved (%) 1.03 (1.01-1.04) .007No. of comorbidities ([3 vsbelow)

9.5 (2.86-31.62) \.001

SCORTEN (on admission) 3.83 (2.02-7.25) \.001SCORTEN (at 48 hours) 2.86 (1.68-4.86) \.001Days to admission 1.04 (0.92-1.17) .52Days in Burn Unit 0.94 (0.88-1.01) .09Metabolic syndrome(yes vs no)

2.88 (0.99-8.40) .05

HIV (yes vs no) 0.96 (0.17-5.35) .97Gout (yes vs no) 6.75 (1.80-25.36) .005Seizures (yes vs no) 0.57 (0.11-2.90) .5Infectious (yes vs no) 0.47 (0.17-1.31) .15

CI, Confidence interval; IVIg, intravenous immunoglobulin.

J AM ACAD DERMATOL

OCTOBER 2012634 Firoz et al

risk of death (OR = 9.24, 95% CI = 2.48-34.47, P\.001). Interestingly, patients with metabolic syn-drome had an almost 3-fold higher risk of death(OR = 2.88, 95% CI = 0.99-8.41, P = .05), although thathas marginal statistical significance, whereas thosewith gout had almost a 7-fold increased risk of death(OR = 6.75, 95% CI = 1.80-25.36, P = .005). Theassociations between risk of death andHIV infection,seizure disorders, or infectious comorbidities werenot statistically significant. Patients with a higherpercentage of body surface area (BSA) involvementhad a higher risk of death; for every 1% increase inBSA, there was a 3% increase in the risk of death(OR = 1.03, 95% CI = 1.007-1.045, P = .007). A 1-unitchange in SCORTEN scores on admission had nearlya 4-fold increase in the risk of death and nearly a3-fold increase for SCORTEN assessed at 48 hours (atadmission: OR = 3.83, 95% CI = 2.02-7.25, P\ .001;and at 48 hours: OR = 2.86, 95% CI = 1.68-4.86, P\.001). Interestingly, treatment with IVIg, the numberof days it took to be transferred to the Burn Unit, orthe overall number of days in the Burn Unit were notassociated with risk of death.

Because certain factors in our analysis were foundto impact mortality, we included them in theSCORTEN statistic to see if it would improveSCORTEN’s ability to predict mortality. After com-bining the presence of gout and metabolic syndromewith the SCORTEN statistic in a logistic regressionanalysis, the area under the ROCmodestly improvedto 0.845 from 0.830. This difference was not statisti-cally different (P = .421), suggesting that adding gout

and metabolic syndrome did not incrementally im-prove the mortality prediction over SCORTEN alone.

CONCLUSIONThis prospective cohort study is one of the largest

groups of TEN patients to be reported from a singleburn unit in just 5 years. We were interested indetermining whether SCORTEN would accuratelypredict mortality in our patient population. Therewere conflicting reports as to its accuracy, especiallyin populations outside Europe.13,14 As shown in thisarticle, SCORTEN accurately predicted mortality andis appropriate for use in burn units or hospitals thatcare for patients with TEN.

An important result is that we found that the use ofIVIg was not significantly associated with improvedsurvival in our cohort. There have beenmixed resultsin the literature; some studies showed improvedsurvival with IVIg use and some did not. One studyfound an increase in mortality with the use of IVIg,14

whereas the other found an 83% reduction ofmortality.13 These differences were attributed todifferent dosing of IVIg (2 g/kg compared with4 g/kg) and timing of drug initiation (3 vs 4 days).All our patients received 4 g/kg of IVIg if indicatedand if blistering of skin occurred 72 hours aroundadmission to the Burn Unit. Despite the earlyinitiation of IVIg at higher doses in our patientpopulation, we still failed to see an improvement insurvival.

When investigating the drugs implicated in theetiology of TEN, we found trimethoprim/

J AM ACAD DERMATOL

VOLUME 67, NUMBER 4Firoz et al 635

sulfamethoxazole followed by anticonvulsants,NSAIDs, and allopurinol to be most commonlyassociated. This corroborates what was found in alarge case-control study of drugs associated withTEN.2 However, if one looks at the mortality associ-ated with each drug, a majority of patients died ifthey were exposed to allopurinol, in contrast to theother drugs on record. This finding could be attrib-uted to several explanations. We hypothesize thatdrug half-life is a very important factor in theprogression of TEN and, ultimately, overall survival.If the offending agent is still circulating in thebloodstream, the cascade of events will not stop.Allopurinol is metabolized quickly within 1 to 3hours; however, its active metabolite, oxypurinol, isexcreted slowly over 18 to 30 hours by the kidney.This is in contrast to trimethoprim/sulfamethoxa-zole, which has a half-life of 8 to 10 hours, oribuprofen, which has a half-life of approximately 2hours.15 A more informative approach would be toprospectively measure actual blood drug levels seri-ally in a cohort of patients, which has yet to be done.

Finally, certain factors were analyzed to see if theyimpacted survival. As is already known, age over 40years, higher SCORTEN, and higher percentage ofinvolved body surface area conferred poorer sur-vival. Higher proportions of patients died if theoffending drug was allopurinol, compared withother medications. Having HIV infection, cancer, ora seizure disorder did not confer increased mortality,although our number of patients with these comor-bidities was small. Interestingly, as the number ofmedical comorbidities increased, and if patients hadmetabolic syndrome or gout, their survival wasworse. Adding gout and metabolic syndrome toSCORTEN modestly improved the ROC curve, al-though it was not statistically significant.

There are a few limitations of this study. The first isthat we had a relatively small sample size because ofthe rarity of TEN. This may have affected the preci-sion of the analyses that were conducted. However,although we had 82 cases, this represents one of thelarger consecutive series of patients that have beenconsecutively assessed. Prior drug exposure washeterogeneous and, combined with the relativelysmall sample size, this impacted our ability to makeinferences about the relationship between any onesingle drug and outcome. Also, because this was aprospective cohort study where the treatment regi-men was chosen by the treating physician, subjectswere not randomized to alternative treatment regi-mens. This increases the potential for bias to haveaffected the comparison between treatment and

outcome. In particular, there were very few patientstreated with cyclosporine.

In conclusion, mortality in our cohort was similarto what has been described in the literature.SCORTEN was confirmed as an accurate predictorof mortality in our cohort. As previously known,higher SCORTEN, older age, and larger BSA involve-ment conferred worse survival. New informationfrom this analysis shows worse mortality if patientshad the presence of additional comorbidities includ-ing metabolic syndrome and gout; the addition ofthese comorbidities to SCORTEN also improved itspredictive ability.

REFERENCES

1. Lyell A. Toxic Epidermal Necrolysis: an eruption resembling

scalding of the skin. Br J Dermatol 1956;68:355-61.

2. Roujeau J-C, Kelly JP, Naldi L, Rzany B, Stern RS, Anderson T,

et al. Medication Use and the Risk of StevenseJohnsonSyndrome or Toxic Epidermal Necrolysis. N Engl J Med 1995;

333:1600-8.

3. Roujeau J-C, Guillaume J-C, Fabre JP, Penso D, Flechet M, Girre

P. Toxic epidermal necrolysis (Lyell syndrome). Incidence and

drug etiology in France, 1981-1985. Arch Dermatol 1990;126:

37-42.

4. Pereira FA, Mudgil AV, Rosmarin DM. Toxic epidermal necrol-

ysis. J Am Acad Dermatol 2007;56:181-200.

5. Bastuji-Garin S, Fouchard N, Bertocchi M, Roujeau JC, Revuz J,

Wolkenstein P. SCORTEN: a severity-of-illness score for toxic

epidermal necrolysis. J Invest Dermatol 2000;115:149-53.

6. Trent JT, Kirsner RS, Romanelli P, Kerdel FA. Use of SCORTEN to

accurately predict mortality in patients with toxic epidermal

necrolysis in the United States. Arch Dermatol 2004;140:890-2.

7. Paul C, Wolkenstein P, Adle H, Wechsler J, Garchon HJ, Revuz J,

et al. Apoptosis as a mechanism of keratinocyte death in toxic

epidermal necrolysis. Br J Dermatol 1996;134:710-4.

8. Wehrli P, Viard I, Bullani R, Tschopp J, French LE. Death

receptors in cutaneous biology and disease. J Invest Dermatol

2000;115:141-8.

9. Hewitt J, Ormerod AD. Toxic epidermal necrolysis treated with

cyclosporin. Clin Exp Dermatol 1992;17:264-5.

10. Arevalo JM, Lorente JA, Gonzalez-Herrada C, Jimenez-Reyes J.

Treatment of toxic epidermal necrolysis with cyclosporin A.

J Trauma 2000;48:473-8.

11. Reese D, Henning JS, Rockers K, Ladd D, Gilson R. Cyclosporine

for SJS/TEN: a case series and review of the literature. Cutis

2011;87:24-9.

12. Prins C, Kerdel FA, Padilla RS, Hunziker T, Chimenti S, Viard I,

et al. Treatment of toxic epidermal necrolysis with high-dose

intravenous immunoglobulins: multicenter retrospective anal-

ysis of 48 consecutive cases. Arch Dermatol 2003;139:26-32.

13. Trent JT, Kirsner RS, Romanelli P, Kerdel FA. Analysis of

intravenous immunoglobulin for the treatment of toxic epi-

dermal necrolysis using SCORTEN: The University of Miami

Experience. Arch Dermatol 2003;139:39-43.

14. Bachot N, Revuz J, Roujeau JC. Intravenous immunoglobulin

treatment for Stevens-Johnson syndrome and toxic epidermal

necrolysis: a prospective noncomparative study showing no

benefit on mortality or progression. Arch Dermatol 2003;139:

33-6.

15. Micromedex 2.0 Accessed July 22, 2011.