Documente Academic
Documente Profesional
Documente Cultură
research-article2014
AOPXXX10.1177/1060028014560012Annals of PharmacotherapyLaw and Leung
Review Article
Annals of Pharmacotherapy
Steroids in Stevens-Johnson
1–8
© The Author(s) 2014
Reprints and permissions:
Syndrome/Toxic Epidermal Necrolysis: sagepub.com/journalsPermissions.nav
DOI: 10.1177/1060028014560012
Future Research
Abstract
Objective:To review the evidence for the use of steroids in adults presenting with Stevens-Johnson Syndrome (SJS),
toxic epidermal necrolysis (TEN), or overlap. Data Sources: EMBASE (1974 to April 2014), MEDLINE (1946 to April
2014), Cochrane Database of Systematic Reviews, and International Pharmaceutical Abstracts (1970 to January 2014) were
searched using the terms: prednisone, methylprednisolone, dexamethasone, prednisolone, steroids, glucocorticoids, corticosteroids,
Stevens-Johnson Syndrome, toxic epidermal necrolysis, and SJS/TEN overlap. Study Selection and Data Extraction: English-
language, full reports of experimental and observational studies were included. Bibliographies from pertinent publications
were reviewed for additional references. Prespecified outcomes included survival, survival to discharge, hospitalization
without intensive care, length of intensive care stay, duration of hospitalization, ophthalmological complications, infection
rates, and adverse events. Data Synthesis: Six studies that used steroids for SJS, TEN, and/or overlap were included. All
studies were retrospective cohort studies with no case-control or cross-sectional studies; 5 studies reported on steroid
doses, and 2 studies reported time from disease onset to steroid use (2-4 days). Only 1 out of 6 studies reported a statistically
significant impact on mortality with steroids use (odds ratio = 0.4; 95% CI = 0.2-0.9). Adverse event rates were not reported
in any of the studies. Conclusions: A review of the current evidence reveals a need for prospective, randomized controlled
studies to provide more definitive conclusions on steroid use in patients with SJS, TEN, and/or overlap.
Keywords
Stevens-Johnson, toxic epidermal necrolysis, steroids
TBSA, whereas SJS/TEN overlap syndrome describes Stevens-Johnson Syndrome, toxic epidermal necrolysis, and
patients with involvement of >10% but <30% of TBSA. SJS/TEN overlap.
The leading cause of SJS and TEN is medication expo-
sure, with a wide range of associated drug classes, including
antibiotics, nonsteroidal anti-inflammatory drugs, psycho-
Study Selection and Data Extraction
tropics (such as antipsychotics and antiepileptic drugs), and This review included any randomized trials, cohort studies,
antigout medications (such as allopurinol).5 However, there and case-control studies in English, with adult human par-
are subsets of cases that are associated with infection or ticipants. Review articles, case reports or series, and cor-
where the definitive cause is unclear. respondence or letters to the editor were excluded.
The prognosis differs greatly across the spectrum of SJS/ Bibliographies from pertinent publications were reviewed
TEN. SJS is usually benign and associated with a mortality for additional references. A manual review of reference
rate of 1% to 3%, whereas in the case of TEN, this rate is as lists from retrieved articles was performed to identify any
high as 25% to 35%.6,7 Therefore, a validated prognostic additional studies. The World Health Organization
scoring system, SCORTEN, was developed based on 7 clin- International Clinical Trials Registry Platform (WHO
ical and laboratory variables to guide assessment of a ICTRP [http://www.who.int/ictrp/]) was also searched for
patient’s severity of illness, need for intensive care, and relevant studies. Studies that evaluated administration of
prognosis (see Appendix).3 steroids to adults with SJS, TENS, or overlap that provided
Management of SJS/TEN involves the early identifica- our defined primary outcome of interest with statistical
tion of the trigger(s) followed by multidisciplinary sup- analysis were considered as meeting inclusion criteria for
portive measures to alleviate symptoms and prevent further this review. Measures of study quality, such as randomiza-
complications of the disease, including wound/burn care, tion, blinding of assessors, prespecified outcomes, and
ocular care, pain control, and prevention/treatment of subgroup analyses were noted. The primary outcome of
infections. There are several pharmacological therapies interest was survival. Other prespecified outcomes of inter-
available, such as steroids, intravenous immunoglobulins est were survival to discharge, hospitalization without
(IVIG), and cyclosporine.5 However, a universally accepted requiring intensive care, length of intensive care stay, dura-
drug treatment regimen has not been established, and much tion of hospitalization, ophthalmological complications,
controversy exists as to the role of each of these drugs indi- infection rates, and any adverse events.
vidually or in combination. In particular, a number of
issues have emerged with the use of steroids in acute man-
agement of these conditions. Much debate remains over the
Data Synthesis
role of steroids because they theoretically increase the risk The literature search yielded 814 citations. On screening
of sepsis, increase protein catabolism, and decrease the rate titles and abstracts, 799 citations were excluded, and 15
of epithelialization; also, other studies have found that potentially relevant articles were retrieved for full-text
administration of systemic steroids was associated with review. No additional potentially relevant reports were
increased morbidity and mortality.8 Moreover, if steroids identified through gray literature searching (ie, WHO
are to be used, the timing, dose, and duration of therapy ICTRP and conference abstracts). Of the 15 potentially rel-
remain unclear. evant reports, 9 did not meet the inclusion criteria. Six pub-
lications were included in this review. No randomized
controlled trials (RCTs) or evidence-based clinical practice
Objectives
guidelines were identified. All studies were retrospective in
The objective of this article is to review the evidence for the nature (Table 1). The study selection process is outlined in a
use of steroids in adults acutely presenting with SJS, TEN, flowchart (Figure 1).
or SJS/TEN overlap. In 2000, Schulz et al9 published the results of a retro-
spective cohort study evaluating the frequency and predic-
tors of mortality among patients admitted to a specialized
Data Sources
burn unit with a diagnosis of SJS, TEN, or erythema multi-
The literature review process is summarized in Figure 1. forme over an 11-year period (1987-1998). Patients included
Published articles were systematically searched from had epidermal sloughing of greater than 20% TBSA. The
EMBASE (1974 to April 2014), MEDLINE (1946 to April primary study objective was the identification of predictors
2014), Cochrane Database of Systematic Reviews, and for in-hospital mortality. The study cohort consisted of 39
International Pharmaceutical Abstracts (1970 to April patients, with almost all cases associated with drug as the
2014). Combinations of the following search terms were etiological agent. Of the cohort, 87.1% received docu-
used: prednisone, methylprednisolone, dexamethasone, mented steroids (21 in those that survived and 13 in those
prednisolone, steroids, glucocorticoids, corticosteroids, that died); 17 (43.6%) patients died while hospitalized.
Through univariate analysis, steroid use was not identi- limitation of this study is the statistical method used by the
fied as a predictor of mortality among those who died and authors because a univariate analysis was likely inappropri-
those who survived. Instead, patients who survived were ate to simultaneously adjust for the numerous potential
typically younger (47.5 ± 4.2 vs 64.5 ± 5.3 years of age, P = confounders.
0.015), admitted earlier after onset of rash (3.5 ± 0.4 vs 5.9 In 2005, Kim et al10 published the results of a retrospec-
± 1 days, P = 0.02), febrile on presentation (101.0 ± 0.5°F tive cohort study of patients hospitalized for the treatment
vs 98.5 ± 0.5°F, P = 0.002), hospitalized for longer (38 ± 6 of TEN in a medical center in Korea from 1990 to 2003.
vs 18 ± 3 days, P = 0.012), less burdened with comorbidi- Patients hospitalized received monotherapy with either cor-
ties (2.5 ± 0.4 vs 4.5 ± 0.8, P = 0.03), and less likely to ticosteroids (IV methylprednisolone 250-1000 mg/d, later
receive pre–burn unit admission antibiotics (43% vs 88%, stepped down to oral prednisolone) or IVIG (1.6-2.0 g/kg).
P = 0.005) or experience thrombocytopenia compared with The primary outcome of mortality was reported as a stan-
those who died. The authors concluded that their results dardized mortality ratio (SMR), which is calculated as the
suggest poorer prognosis with regard to mortality in patients observed frequency of deaths over the expected frequency
with these characteristics. of death (based on the SCORTEN prognostic score). A total
A strength of this study was that it provided evidence of of 38 patients (20 male, 18 female) were identified, with the
the complexities and potential prognostic characteristics most common etiology of TEN being antibiotic exposure
that may influence the course of SJS/TEN. The greatest (18 cases, 47.4%), followed by infection (14 cases, 36.8%).
Abbreviations: GC, glucocorticoid; HR, hazard ratio; IVIG, intravenous immunoglobulins; NR, not reported; NS, not significant; SJS, Stevens-Johnson syndrome; SMR,
standardized mortality ratio; TBSA, total body surface area; TEN, toxic epidermal necrolysis.
The average TBSA involvement ranged between 15% and The strengths of this study are the large cohort study,
70%, with the mean TBSA = 47.7% ± 17.1%. In all, 21 detailed description of treatments, and adjustment for
patients received corticosteroid treatment, and 14 patients potential confounders. However, unmeasured confounders
received IVIG. The remaining 3 patients’ treatments were may still exist given the retrospective, observational design.
not described. This is further confounded with the addition of the 40
No statistically significant differences were noted with patients who received both IVIG and steroids into the treat-
respect to mortality in the overall cohort compared with the ment arms.
SCORTEN-predicted rates of death (SMR = 0.928; 95% CI = In 2009, Yang et al12 published the results of a retrospec-
0.424-1.761). There was no statistical difference in mortality tive cohort study of patients hospitalized in an intensive
in patients treated with steroids (SMR = 1.004; 95% CI = care unit for the treatment of SJS or TEN, in a tertiary care
0.369-2.187). In patients treated with IVIG, investigators did center in China, from 1993 to 2007. Consecutive patients
note that the SCORTEN-predicted mortality rate was 16.8% hospitalized from 1993 to 2000 received corticosteroids,
(2.53 deaths) compared with the observed rate (1 death, whereas those hospitalized between 2001 and 2007 received
7.1%); however, the number of patients was too small to draw combination therapy with IVIG and steroids. The outcome
any final conclusions. The authors concluded that “further of interest was mortality, reported as the SMR based on
large randomized, placebo-controlled trials” were required to SCORTEN-predicted rates of death. Additional outcomes
evaluate the consequences of treating TEN with IVIG. included time to the arrest of progression, time to the taper-
This study adds to the knowledge of the effects of mono- ing of corticosteroids, time of hospitalization, side effects,
therapy steroids or IVIG in treating patients with TEN in and complications. A total of 65 patients (18 cases of SJS
comparison with predicted mortality rates. Limitations of and 47 cases of TEN) were identified, with no statistically
this study include the retrospective design and nonrandom- significant differences between actual (13 deaths, 20%) and
ized selection of patients, with an absence of adjustment for SCORTEN-predicted (12.14 deaths, 18.7%) mortality rates.
potential confounders. The most common etiology for SJS or TEN was docu-
The largest study to evaluate the effect of treatments, mented to be drug-related, and the average TBSA involve-
including steroids, was performed by Schneck et al.11 The ment ranged between 15% and 70%, with the mean TBSA
authors examined data procured from a case-control study = 41% ± 11%. In all, 45 patients received corticosteroid
that evaluated the risk factors for SJS or TEN in 6 countries treatment, and 20 patients received combination therapy.
in Europe (Austria, France, Germany, Israel, Italy, and the No statistically significant differences were noted with
Netherlands). The primary outcome was death during hos- respect to mortality in the overall cohort (SMR = 1.16;
pitalization. Of a cohort of 379 patients with confirmed SJS 95% CI = 0.56-2.13). This non–statistically significant
and TEN, 281 patients with information about treatment difference in mortality was consistent in both SJS and
were identified in 4 different treatment groups (supportive TEN subgroups. In patients with TEN, investigators did
care, n = 87; IVIG only, n = 35; IVIG + steroids, n = 40; and find a statistically significant difference in mean time to
steroids only, n = 119). Because of the small number of par- arrest of progression (P = 0.0188) and total hospitalization
ticipants in the combination group (IVIG + steroids), the time (P = 0.0034). In patients with SJS, a statistically sig-
effects of this arm were not estimated. Logistic regression nificant reduction in mean time to arrest of progression
was used in the comparisons to generate odds ratios (OR). (P = 0.019) and total hospitalization time (P = 0.0475)
Of the 159 patients who received any steroids, 28 were observed. No other statistically significant differ-
(17.6%) died during hospitalization. Compared with sup- ences were seen in the other reported outcomes and sub-
portive care, there was a statistically significant decrease in groups. The authors concluded that combination therapy
the odds of death (OR = 0.4; 95% CI = 0.2-0.9) after multi- with steroids and IVIG had a “tendency” to reduce mortal-
variate adjustment for age group (<40, 40-70, >70 years), ity, and though this result was not statistically significant,
disease severity (SJS, SJS/TEN, TEN), and country. Agents the positive results in the secondary outcomes support the
used in the steroid-exposed patients were prednisone, meth- use of combination therapy.
ylprednisolone, and dexamethasone, started on admission This study adds to the knowledge of managing patients
after a mean of 4 (2-5) days after disease onset at a mean with steroid therapy compared with combination treatment.
dose of 250 (100-500) mg/d of prednisone for a mean dura- However, there are major limitations of this study such as
tion of 4 (2-12) days. The authors conclude that there was the retrospective design and nonrandomized selection of
inadequate evidence that any specific treatment is estab- patients, without adjustment for potential confounders. In
lished as effective for patients with SJS or TEN, with only particular, the period of time for which the steroid group
corticosteroids showing a “trend for possible benefit.” They was recruited (1993-2000) may differ significantly from the
call for a prospective randomized trial to be conducted combination therapy group (2001-2007) with respect to
before any conclusions can be drawn, recommending that supportive care measures and other therapies. Finally, the
corticosteroids be trialed first. authors report P values for impact of the treatments on
mortality but do not provide event rates, point estimates, or varying equivalent doses of prednisone depending on geo-
CIs to accompany these. graphical location (France, 75 mg; Germany, 250 mg).
In a study of patient data collected from a South Korean Time to start of therapy after disease onset and duration of
academic tertiary care hospital during the period 2001- treatment was not reported. The statistical methods for
2011, Kim et al13 characterized risk factors for mortality assessing treatment effect on survival were not clearly
among patients with a diagnosis of SJS or TEN (n = 82). reported by the authors, but survival was not changed by
Patient data and information related to treatment and out- steroid use (hazard ratio = 1.3; 95% CI = 0.8-1.9).
comes were collected from medical records and analyzed The large population-based cohort with time-adjusted
by multivariate logistic regression to determine ORs for survival analysis is a strength of this study. However, the
mortality among several potential covariates for mortality. lack of clarity regarding statistical methods and the “any
Of the 82 cases, 52.4% were considered drug-related etiolo- exposure” definition of corticosteroid use results in diffi-
gies and predominantly single-drug causes. Most patients culty in gaining insight into the incremental effect of ste-
(70/82, 85.4%) received IV steroids, and most of them were roids in these patients.
given dexamethasone (52/70, 63.4%) at a dose of 10 mg/d,
started after a mean of 2.21 (1-15) days after admission ini-
Discussion
tially. Oral steroids (predominantly prednisolone 15 mg/d)
were used in 42 (51.1%) patients during their hospitaliza- Corticosteroids are a class of medications that have been
tion. The mean duration for IV and PO corticosteroid use long used for their immunosuppressive and anti-inflammatory
was 12.68 days (1-68 days). In all, 8 patients died during effects for treatment of a wide spectrum of disorders such as
hospitalization (9.8%). rheumatoid arthritis and inflammatory bowel disease.15 It
Through univariate analysis, steroid use was not identi- has been suggested that the first known use of steroids for
fied as a predictor of mortality among those who died and the treatment of SJS was reported in 1951 by Bleier and
those who survived and, therefore, not included in the sub- Schwartz16 in several cases of SJS with severe ocular mani-
sequent multivariate analysis. The only statistically signifi- festations. It was reported that these patients achieved sig-
cant predictor of mortality was the diagnosis of pneumonia nificant improvement in their symptoms with the
(OR = 25.79; 95% CI = 2.25-296.21; P = 0.009). The administration of cortisone or adrenocorticotropic hor-
authors concluded that the low mortality rate compared mone.17 As more reports of success with corticosteroids
with previous studies may be a result of the high rates of were published, the standard of treatment for children, ado-
steroid use, and therefore, suggested that steroid therapy lescents, and adults presenting with these conditions
should be used early, with careful documentation of signs included the use of corticosteroids. Controversy around this
and symptoms of upper-respiratory-tract infection. practice first arose when 3 retrospective case review studies
The provision of detailed information characterizing ste- were published.18-21 These studies challenged the established
roid use and other potential prognostic indicators strength- standard of care that corticosteroids shorten the course of the
ens the study. However, it would be premature to draw any disease and improve survival rates. Instead, these authors
firm conclusions about the perceived benefit of early ste- proposed that exposure to these agents increase the risk of
roid use, given that no statistically significant differences infectious complications and prolong hospitalization.
were found. These results may have been a result of type II Our review of the data to date, which excluded low-quality
error, based on the low mortality rate and the small size of case reports, case-series, and reviews, but is limited to out-
the study cohort. come data from retrospective cohort studies, suggests that
In 2013, Sekula et al14 published the results of a retro- the impact that steroids have on mortality among patients
spective cohort study based on patients with a diagnosis of presenting with SJS, SJS/TEN, and TEN is inconclusive.
SJS, SJS/TEN overlap, and TEN enrolled in the multina- Despite our attempts to include the highest-quality studies
tional, RegiSCAR study (2003-2007). The authors con- possible, this review is still limited by several reoccurring
ducted several analyses of the risk of mortality based on weaknesses of the included studies. All studies are limited
age, severity of reaction, organ function, pertinent comor- by their nonrandomized design, which increases their sus-
bidities/infections, and treatments received. The treatment ceptibility to bias and confounding. There are also inconsis-
analysis included 442 cases, with 97 (22%) receiving sup- tencies in the reporting of disease severity and characteristics
portive care only; 317 (72%) were exposed to steroids, 81 of supportive care that create difficulty in applying the
(18%) exposed to IVIG, and 24 (5%) received another study results to clinical practice. Furthermore, several out-
immunomodulating drug (4.2% were given cyclosporine). comes of interest were not reported in many of the studies,
Also, 244 patients (55%) received corticosteroids only, but particularly adverse events. Finally, the small study sizes
this was not included in the statistical analysis for determin- and differences in disease severity between those who
ing risk of mortality. Patients were admitted to hospital for received steroids and those who did not introduces type II
a median time of 2 days (range = 0-23 days) and received error and confounding.