Pediatric SJS-TEN: Where are we now?

Stevens–Johnson syndrome and toxic epidermal necrolysis are rare severe blistering skin reactions triggered by medications or infections. Over the last 5 to 10 years, a number of important publications have advanced understanding of these diseases and their response to treatment. Importantly, a subset of patients with disease triggered by infection has been identified as having Mycoplasma pneumoniae–induced rash and mucositis, suggesting a reconsideration of the diagnostic paradigm. We present an update on pediatric Stevens–Johnson syndrome and toxic epidermal necrolysis in the broader context of cutaneous adverse drug reactions and focus on challenges and recent advances in diagnosis, management, and prevention.


Introduction
Skin (cutaneous) reactions are among the most common types of adverse reactions to medications in children, accounting for 36% of any adverse drug reaction (ADR) 1 . Many of these reactions are mild and self-resolving, but the rare severe ADRs can be associated with significant morbidity and even mortality. The diagnosis and treatment of cutaneous ADRs (cADRs) in children are challenging for several reasons. First, children are more commonly infected with viruses as compared with adults; each year, children in their first years of life average six to 10 respiratory viral infections and older children and adolescents average three to five such illnesses 2 . Many of these viruses, such as Epstein-Barr virus, adenovirus, and enteroviruses, are frequently associated with cutaneous reactions that can be misinterpreted as cADRs. Second, triggers of severe cADRs, such as Stevens-Johnson syndrome (SJS), are different in adults and children; medications more often are implicated in adults and infections more commonly cause SJS in children 3 . Third, causality tools currently available-for example, Naranjo, algorithm of drug causality for epidermal necrolysis (ALDEN), and the US Food and Drug Administration's division of drug experience-have been developed and studied primarily in adults and not applied to children. Finally, owing to the overall rarity of severe cADRs in children, there is little evidence to guide treatment.
ADRs can be either type A (predictable, caused by on-target drug action) or type B (heterogeneous, immunologically and metabolically mediated) (Table 1). cADRs can be classified clinically by morphology and severity on the basis of the presence (complex) or absence (simple) of fever and other systemic symptoms (Table 2).
Given their rarity, complex or severe cADRs are difficult to study systematically and pediatric-specific data are limited. These challenges have led to the adoption of adult paradigms for diagnosis and management in pediatric practice, where better evidence exists.
In SJS and toxic epidermal necrolysis (TEN), severe blistering of the skin and mucous membranes related to either medications or infections occurs. The differential diagnosis in early cases where diffuse erythema is common may include viral exanthems, Kawasaki disease, and acute generalized exanthematous pustulosis. As widespread blistering begins to develop, thermal burns, toxic erythema of chemotherapy, pemphigus of all types, staphylococcal scalded skin syndrome (which spares the mucous membranes), acute graft-versus-host disease, acute syndrome of apoptotic pan-epidermolysis associated with systemic lupus erythematosus, and generalized bullous fixed drug eruption become the primary considerations.
In recent years, significant developments have occurred in the category of severe blistering drug reactions-SJS and TEN-the focus of this review. The watershed evolutions in this area will be reviewed and ultimately support the need  for evolution of the diagnostic categories for severe blistering cADRs in children.

Incidence
Until 2017, estimates of the incidence of SJS and TEN were based on retrospective surveys and databases from the 1990s and a European registry 6-9 . More recently, larger datasets have been used in attempts to better describe these rare events in pediatrics.
A recent US pediatric database cohort study suggests that SJS-TEN spectrum disorders occur in 7.5 per 100,000 hospitalized children 10 ; incidences of 6.3 and 0.5 per 100,000 hospitalized children per year for SJS and TEN, respectively, were reported. A second US cross-sectional study that sourced data from the same 2009-2012 time frame found much lower rates, of 5.3 (SJS) and 0.4 (TEN) per million children per year, in the overall population 11 . Similar findings in both studies suggest that pediatric SJS-TEN results in a substantial health-care burden, although mortality is less when compared with adult data. There are many limitations to using large datasets to examine rare events such as SJS-TEN, including poor validity of International Classification of Diseases 9/10 (ICD-9/10) codes, inability to determine cause, and the lack of a standardized diagnostic approach across institutions. Given that the clinical features of severe cADRs can be challenging to interpret for diagnosis, standardization in case identification and validation is needed.

Clinical features and refining diagnosis: drug versus bug
Stevens-Johnson syndrome and toxic epidermal necrolysis SJS and TEN are characterized by blistering of the skin and mucous membranes. One to three days before onset of skin and mucosal lesions, prodromal symptoms start; these include fever, general malaise, non-productive cough, stinging eyes, and a sore mouth. These symptoms are often mistaken for an upper respiratory tract infection. Macules with purpuric, non-blanching centers with a predilection for the head and torso evolve quickly, often within 12 hours, into blisters that slough off, leaving large areas of denuded skin and mucosa. Painful erythema of the palms and soles is also common early in the disease. Target papules with three distinct rings, characteristic of erythema multiforme, a similar condition usually triggered by herpes simplex virus 12 , are not the main morphology in SJS and TEN, although atypical two-ringed or macular (flat) targets may occur early, before blistering starts. Mucosal involvement affects oral, ocular, genitourinary, and anal sites.
SJS and TEN are believed to exist on a spectrum; there is less than 10% body surface area (BSA) involvement in SJS and greater than 30% BSA involvement in TEN, and intermediate BSA involvement of 10 to 30% is called SJS-TEN overlap (Table 3) 13 . Involved areas include skin that is already blistered or detached and skin that is red (macular erythema) and detachable.
These diagnostic categories were developed in 1993 on the basis of an expert review and synthesis of hundreds of adult cases 13 , and for many pediatric cases, a diagnostic category can be assigned. However, problems arise when patients have severe involvement of their mucous membranes but little or no skin lesions because there is no diagnostic category for them.

Mycoplasma pneumoniae-induced rash and mucositis
In 2015, Canavan et al. 14 described Mycoplasma pneumoniae (MP)-induced rash and mucositis (MIRM) as an entity distinct from SJS. Patients with MIRM have exactly the features of the previously noted patients who evaded classification; they have severe mucositis of multiple mucous membranes out of proportion to skin involvement, which typically is sparse but in some cases may be significant 14 . The characteristic constellation of features in MIRM is triggered by respiratory infection rather than medications, and pathogens other than MP have been reported 15 . Advances in technology, including respiratory polymerase chain reaction (PCR), have improved MP detection, and more specific methods to confirm infection have recently been developed 16 ; these include measurement of MP-specific antibody-secreting cells 17 .
The challenge with cases attributed to infection and with the diagnosis of MIRM is that their clinical features overlap with those of SJS, creating opportunities for patients to receive multiple diagnoses and complicating more comprehensive and systematic study of these cases in the future. Furthermore, as there is no diagnostic code in the ICD-9 or ICD-10 for MIRM, these patients would likely have been, and continue to be, assigned codes for SJS, TEN, or erythema multiforme with unreliable assignment with secondary codes for mycoplasma infection.

Proposed revised classification
Author MLR is part of a group that recently proposed a revised classification for severe blistering cutaneous reactions in children; the revision condenses SJS and TEN into a single category of drug-induced epidermal necrolysis (DEN) that may have variable skin involvement (manuscript under revision). This is logical because SJS and TEN are considered quantity variants of the same disease and drugs are their common trigger.
The proposed classification separates out the infection-related cases typified by severe mucosal and less impressive skin lesions as reactive infectious mucocutaneous eruption (RIME).
Given their disparate causes and pathogeneses, DEN and RIME have diverging management strategies: the focus is on drug withdrawal and early consideration of immunosuppressive treatment in patients with DEN and on identification and treatment of infection combined with supportive care with or without immunosuppressive therapy in RIME cases. Application of this novel pediatric-specific paradigm may have far-reaching impacts on incidence and epidemiology and ultimately provide more directed and effective management.
Biomarkers will likely emerge as tools to confirm diagnosis; the granulysin rapid test, which has a turnaround time of 15 minutes, was used in a 2011 report to establish an early diagnosis of SJS in a young child 18 . Other candidate biomarkers, including eosinophilia, perforin, interferon-gamma, soluble Fas ligand, and CD69, are being evaluated, although none is validated for clinical use at this time 19 .

Etiologies
Recent pediatric retrospective studies identify a drug as the cause of SJS and TEN in 72 to 90% of cases 20-23 . Idiopathic cases, where no cause (medication or infection) can be identified, make up 5 to 17% of cases in retrospective reviews 20,21,24 . A limited number of medications, including antibiotics, anticonvulsants, and non-steroidal anti-inflammatory drugs, are implicated in the majority of pediatric cases. As stated above, MP is a common trigger of severe mucocutaneous reactions that have been considered SJS-TEN in the past; however, the clinical presentation and outcomes suggest a unique disease process.

Causality and risk reduction
It is important to avoid implicating medications used to treat the prodromal symptoms of SJS and TEN as causative agents 19 . Survivors of SJS and TEN are often told to avoid all medications taken just prior to the reaction, limiting treatments that can be used for future illnesses 25 and causing patients and families concern about the use of medications in general. Evaluation of possible implicated medications requires utilization of an effective causality tool. Many available ADR causality tools, including the Naranjo and Liverpool tools, are non-specific to the ADR phenotype. The ALDEN is specific for severe cADRs. These tools are helpful when considering the timing of drug exposure to reaction onset, the probability of drug presence in the body, prior exposure to the same drug regardless of reaction at that time, the presence of drug beyond the progression phase, drug notoriety as an ADR cause, and the presence or absence of other etiologies 25 . Although these tools can be helpful, their usefulness is limited by variability in results between users and testing methods. Further advancements in diagnostics and assessment tools have the potential to enhance causality assessment in the future 26 .
Testing to identify the causative agent in SJS-TEN remains controversial and is not widely used. In vivo (varied methods of re-exposure of the patient to potential trigger medications) testing includes patch testing or delayed intradermal testing. These tests can be performed once the acute reaction has resolved but within a year. Potentially cross-reacting medications can also be identified with patch testing 27 . Unfortunately, the results are not reliable for all medications, limiting their clinical utility. Oral re-challenge, recommended for other types of drug reactions, is not recommended for SJS and TEN given the serious risk of a second potentially fatal episode 28 . In vitro testing with the lymphocyte transformation test (LTT) during the recovery phase, 4 to 8 weeks after the reaction, is controversial as LTT historically has had low sensitivity in SJS-TEN and many false-positive and -negative results 29 . Recent reports support better sensitivity (86%) and specificity (74%) in SJS-TEN, even for low-risk drugs 30 . The drug-specific interferon-gamma-releasing cell assay is highly specific (95% specificity and 79% sensitivity for allopurinol) and can be carried out in the acute phase 31 . A recent publication highlighted its advantages over LTT: higher rates of causative drug identification (73.9% versus 52.2% for LTT) and use during the acute phase (versus recovery phase for LTT) 32 .
Prevention of SJS and TEN is becoming a reality with the identification of risk factors that can be screened prior to drug initiation. Human leukocyte antigen (HLA) screening for ethnicity-specific risk alleles prior to administration of aromatic anticonvulsants (including carbamazepine), allopurinol, and abacavir reduces the risk of SJS-TEN (Table 4) 33 .
Polymorphisms that reduce drug clearance can also increase the risk of severe cADRs 34 When a risk allele is identified in a patient, the medication should be avoided 35 . As a proof of concept, in Taiwan, where pre-carbamazepine screening for HLA-B*1502 was adopted in 2010, the incidence of SJS-TEN has drastically decreased 36 . In Thailand, where HLA-B*1502 pre-treatment screening is also routine, HLA test results are printed on a wallet card that patients can carry with them as a "pharmacogenomic ID card" for future health-care interactions 35 .
To treat or not to treat, that is the question… but we still don't know the answer SCORe of Toxic Epidermal Necrolysis (SCORTEN), a TEN severity-of-illness score 37 , was developed for adults but has been used to predict both morbidity and mortality in pediatric patients 38,39 . The SCORTEN has multiple factors, but three of them-age of at least 40 years, malignancy, and heart rate of at least 120 beats per minute-are less important in children because (1) they are young, (2) they have low rates of malignancy, and (3) their normal heart rate can be over 120 beats per minute if they are younger than four. The SCORTEN, or a pediatric-adapted scoring system, could be used prospectively in studies to predict treatment efficacy by comparing actual with predicted mortality and morbidity 38 . Despite the availability of a validated severity score that helps to predict which patients will have more severe outcomes, the absence of a gold-standard effective therapy for pediatric SJS-TEN means that the decision between supportive care and directed (immunosuppressive) therapy remains challenging.
In the last two years, there has been an explosion of retrospective and database publications on pediatric SJS-TEN which have come to similar conclusions on the relative lack of efficacy of corticosteroids and intravenous immunoglobulin (IVIG) on various outcomes 20,22,[40][41][42][43] . For treatment of SJS-TEN, there is a consensus on the importance of rapid identification and withdrawal of the causative medication. As adjuvant therapy, corticosteroids are most frequently used, followed by IVIG both as monotherapy and in combination with corticosteroids. None of these mono-or combination therapies appears to affect time to healing or length of hospital stay. The best evidence on treatment before 2017-2018 was from a large (n = 128 cases) systematic review published in 2011 that suggested that patients who received IVIG and prednisone had better outcomes than those who received supportive care only 44 .
Recent case reports and case series document the rapid efficacy of cyclosporine 3 mg/kg per day divided twice a day and infliximab 5 mg/kg intravenously for one dose in both drug-and infection-related cases 45,46 . There is emerging evidence from case reports for anti-tumor necrosis factor (anti-TNF) therapy in the acute phase [47][48][49][50] . Furthermore, a recent Taiwanese randomized clinical trial of etanercept in TEN included children older than 4 years of age and showed that etanercept (25 mg twice weekly <65 kg and 50 mg if >65 kg) decreased the predicted mortality rate and reduced skin healing time compared with corticosteroids (1 to 1.5 mg/kg per day intravenously) in the group as a whole 51 . Both groups received treatment until their skin lesions healed 51 . This study included five children who were 6 to 13 years old 51 . Further large-scale studies are needed to confirm these promising results in the pediatric population.
Ocular complications and sequelae are significant for pediatric patients with SJS-TEN. Aggressive initial management, including adjuvant amniotic membrane transplantation, can reduce complications 52,53 . Even in the absence of severe ocular involvement in acute SJS-TEN, children may develop progressive disease of the ocular surface and conjunctival inflammation over time 54 . Delayed lid margin keratinization 55 , vision deterioration, and corneal damage can occur after SJS-TEN 56 , emphasizing the importance of close continual follow-up by ophthalmology 55,57 .
The early introduction of psychological and social support is critical for pediatric patients with SJS-TEN to avoid long-term anxiety and depression. Ideally, the introduction should occur as soon as the patient is stable enough for a psychologist or child-life worker to visit. For children, the experience of complete loss of control of their body can be devastating, and early explanation and attention to the patient's emotional needs are critical. Other chronic sequelae are summarized in Table 5.

Treatment of reactive infectious disease: Mycoplasma pneumoniae and other infections
Given the limited data on reactive infectious mucocutaneous eruptions, including MIRM and non-MP, it remains unclear whether treatment should be similar to that of SJS-TEN. A systematic review by Canavan et al. suggests that the majority of patients receive antibiotics and a smaller percentage receive corticosteroids, IVIGs, or supportive care only 14 . Although antibiotics are used to treat MP respiratory infections, the effect of antibiotics on the skin and mucous membrane changes of MIRM is unclear 14 . Recurrent disease is reported in up to 20% of patients 23 , suggesting a genetic susceptibility to reactive mucositis and rash which has yet to be elucidated.
MP-related SJS-TEN overlap has been reported in one pediatric patient to have rapidly re-epithelialized with cyclosporine 3 mg/kg per day over a period of 7 days without co-treatment for MP infection 45 . A second case report of TEN induced by MP responded rapidly with complete resolution within a week after a single dose of infliximab 5 mg/kg per day (the patient also received meropenem) 46 . A case series of three children whose MIRM was treated with cyclosporine suggests that the addition of cyclosporine 3 to 5 mg/kg per day for early cases may accelerate resolution compared with supportive care and antibiotics alone 58 .
Generally, the course of MIRM is less severe than for drug-related SJS-TEN, and supportive care is a reasonable option. However, the mucous membranes should be closely monitored as the same sequelae as are seen in SJS-TEN occur and can be severe. A recent publication in the ophthalmology literature suggests that these children should be followed closely during their inpatient admission similarly to SJS-TEN 59 . In their systematic review, Canavan et al. reported ocular sequelae in 9%, post-inflammatory dyspigmentation in 6%, and oral or genital synechiae in less than 1% 14 . Severe gynecologic sequelae requiring surgery have been reported 60 . Given the sequelae and the acute severe disease course, psychosocial support should ideally be introduced to patients and families at the time of diagnosis (as for SJS-TEN) and there should be a regular review for signs of (latent) distress at follow-up visits to their primary care provider.
The bottom line for therapy is that the identified triggering medication should be discontinued, an infection should be sought and treated if suspected, supportive care in a low-ratio nursing environment should be provided, and anti-inflammatory immunosuppressive therapy, particularly in the early acute phase, should be considered. Specialist input should be sought and coordinated. An overview of the approach to practical management of pediatric SJS-TEN is presented in Table 6 and is detailed by McPherson et al. 61 .

How redefining diagnosis of pediatric Stevens-Johnson syndrome -toxic epidermal necrolysis can improve treatment
The literature on the treatment of pediatric SJS and TEN is disappointing as larger studies suggest no impact of any intervention other than drug withdrawal. In contrast, case to respond similarly to a given therapy? Considering a shift in diagnostic paradigm as discussed above may be a way to interpret the existing literature with a lens focused from cause/trigger to treatment. Adding to this a precision medicine approach that compares subpopulations of responders with non-responders to identify characteristics that permit early recognition and takes advantage of biomarkers like the granulysin rapid test may lead to a future in which we can select the right treatment for the right patient every time.

Conclusions
The future is bright for pediatric SJS-TEN. Initiatives are under way to improve our understanding of this spectrum of disorders, specifically in children. A British guideline on management was recently published 61 . With increasing accessibility and validation of risk factor screening (HLA and metabolism variants) and biomarkers, we may soon be able to prevent SJS-TEN in predisposed individuals and diagnose and treat it early when it occurs by chance. Give patient a wallet card that identifies their history of SJS-TEN and HLA screening result for future medical encounters. Please refer to Figure 4 in Sukasem et al. 35 for an example.

Open Peer Review Current Peer Review Status:
Editorial Note on the Review Process Faculty Reviews are review articles written by the prestigious Members of Faculty Opinions. The articles are commissioned and peer reviewed before publication to ensure that the final, published version is comprehensive and accessible. The reviewers who approved the final version are listed with their names and affiliations.