An individual patient data meta‐analysis of wound care in patients with toxic epidermal necrolysis

Toxic epidermal necrolysis (TEN) involves extensive mucocutaneous loss, and care is supportive. The approach to wound care includes surgical debridement or using dressings while leaving the epidermis intact. Robust evidence for either approach is lacking. We compared surgical debridement to the use of dressings while leaving the epidermis in situ (referred to hereon as dressings) in adult patients with TEN. The primary outcome assessed was mortality. The secondary outcome was time to re‐epithelialisation. The impact of medications was evaluated. An individual patient data (IPD) systematic review and meta‐analysis was undertaken. A random effects meta‐analysis and survival analysis for IPD data examined mortality, re‐epithelisation time and the effect of systemic medications. The quality of evidence was rated per the Grading of Recommendations Assessment, Development and Evaluation (GRADE). PROSPERO: CRD42021266611 Fifty‐four studies involving 227 patients were included in the systematic review and meta‐analysis, with a GRADE from very low to moderate. There was no difference in survival in patients who had surgical debridement or dressings (univariate: p = 0.91, multivariate: p = 0.31). Patients who received dressings re‐epithelialised faster than patients who underwent debridement (multivariate HR: 1.96 [1.09–3.51], p = 0.023). Intravenous immunoglobulin (univariate HR: 0.21 [0.09–0.45], p < 0.001; multivariate HR: 0.22 [0.09–0.53], p < 0.001) and cyclosporin significantly reduced mortality (univariate HR: 0.09 [0.01–0.96], p = 0.046; multivariate HR: 0.06 [0.01–0.73], p = 0.028) irrespective of the wound care. This study supports the expert consensus of the dermatology hospitalists, that wound care in patients with TEN should be supportive with the epidermis left intact and supported with dressings, which leads to faster re‐epithelialisation.


INTRODUCTION Background
Toxic epidermal necrolysis (TEN) is a rare life-threatening condition characterised by severe mucocutaneous losses.It exists on a disease continuum with Stevens-Johnson syndrome (SJS); SJS has less than 10% body surface area involvement, and TEN is typically distinguished by epidermal detachment ≥30% total body surface area (TBSA).][4] TEN is associated with complications including sepsis, [5][6][7] and mortality rates are reported between 14% and 30% [8][9][10] which is strongly dependent on the extent of epithelial loss. 8,9atients with TEN are assessed by the Severity-of-Illness Score for TEN (SCORTEN), a tool for predicting mortality. 11Higher scores are associated with higher mortality, 11 and SCORTEN is widely used to measure and compare outcomes of therapeutic interventions in patients with TEN. 12,13anagement of TEN includes immediate discontinuation of the offending medication, and supportive care.][18][19] The principles of wound care include protection of the underlying exposed dermis, minimisation of infection, reducing the risk of pigment changes and scarring, optimising re-epithelialisation and prevention of insensible fluid losses. 20However, there remain differences in opinion with respect to specific wound care for patients with TEN. 20,21Specifically, wound care can be a surgical approach, that being debridement, or supportive dressings, 17 where the detached epidermis is left in situ to act as a biological dressing for the underlying dermis. 22espite recent guidelines for the management of SJS/ TEN developed from a Delphi consensus of dermatology hospitalists recommending against debridement of necrotic skin, 23 there is limited evidence and no gold or universally accepted standard for wound care in patients with SJS/TEN.Further, there are a lack of randomised controlled studies guiding wound management.As such, inconsistent approaches to wound care continue to exist as seen in the United States, where only 54% of dermatology departments and 61% of burns units having established treatment guidelines for SJS/TEN management. 24ebridement involves the removal of detached, necrotic epidermis followed by the application of suitable dressings and/or antiseptic agents, to aid re-epithelisation. 17,25,26ebridement has been shown to reduce mortality and is adopted as an approach for wound care in patients with TEN in some centres. 17,18,27owever, debridement is often combined with adjuvant therapies, and it is difficult to ascertain that debridement alone contributes to lower infection rates and reduced mortality. 18,28Despite this, guidelines support debridement for TEN patients with sepsis, progressive clinical deterioration and delayed healing. 22,29he alternative involves leaving the epidermis in situ to serve as a biological dressing, in combination with supportive dressings. 16,20,30This has been reported to improve wound healing and has comparable mortality rates to surgical debridement. 16Guidelines from France support a dressingbased approach with minimal manipulation of the skin and non-adhesive dressings, 16 which has been postulated to facilitate early wound healing. 31This approach avoids scarring and hyperpigmentation that can occur following debridement where there can be disruption to the dermis. 15,32

Rationale
Currently, there is no standard approach for wound care in patients with SJS/TENS.7][18][19] However, there are no standardised protocols, and the wound management between units is variable. 10,22,33While expert consensus favours a non-surgical approach, 34 there are a lack of randomised controlled trials guiding optimal wound care in patients with SJS/TEN.As such, establishing standardised wound management in patients with TEN remains a significant, unmet clinical priority.

Objectives
The primary objective was to determine whether surgical debridement compared to dressings will lead to a mortality benefit in patients with TEN.
Secondary objectives involved assessing the time to reepithelialisation, and the effect of systemic therapies on mortality and re-epithelialisation.

Study design
The study protocol was registered with PROSPERO (CRD42021266611) on 14 July 2021.The findings were reported according to the PRISMA-IPD 2015 guidelines. 34An online database search for all studies addressing wound care in patients with TEN was undertaken.Individual patient data were extracted including demographics (age, sex, country, comorbidities, cause, TBSA and other organ involvement), SCORTEN, intervention (dressing or debridement), mortality, re-epithelisation time, sequelae and adjuvant therapies.

Search strategy and eligibility criteria
A systematic search of PubMed, SCOPUS and Cochrane Library databases was performed for relevant studies published from 1 January 2000 to 28 February 2022.The following search terms were used: 'toxic epidermal necrolysis' combined with 'debridement' or 'dressing' or 'management'.Reference lists of relevant publications were searched for additional studies meeting the inclusion criteria.
Studies were included if (1) patients with TEN had their wounds managed with surgical debridement or dressings; (2) patients were aged 18 years and older; (3) English-written articles.In addition, case reports, case series and cohort studies were included due to the low number of studies available in the literature reporting on patient outcomes.Studies were excluded if they did not report outcomes or contained missing patient data that could not be obtained despite contacting the authors.Exclusion criteria were as follows: (1) articles with duplicate patients; and (2) articles in a language other than English.

Study selection
The articles were selected using a two-stage process by two reviewers (JL and RBS).All titles and abstracts were reviewed independently, and full texts of articles were reviewed if inclusion was unclear.Any discrepancies/ disagreements were resolved by discussion with a third reviewer as indicated (KM).Detailed information regarding the study selection process is provided in the PRISMA flow diagram (Figure 1).

Data collection and data items
Data were extracted (JL) and checked for completeness and accuracy by RBS.The following information was extracted: characteristics of study (citation, study period and design), participant characteristics (age, sex, %TBSA and SCORTEN), wound care approach (dressing or debridement), death, co-morbidities, cause, dressing type, adjuvant treatments (intravenous immunoglobulin [IVIG], cyclosporin, oral corticosteroids and etanercept), time to re-epithelisation, length of hospital stay and morbidity outcomes where available (such as scarring, ophthalmological sequelae and sepsis).Individual participants were allocated to the 'dressing' group if they did not undergo any debridement.Similarly, participants who underwent any debridement were allocated to the 'debridement' group.These debrided participants represented centres practising conservative wound care [35][36][37][38][39][40][41][42] and centres where routine debridement is practised. 29,32,43,44,45,46,47,48,49,50,51,52,53,54,55here the data were not available in the article, the study authors were contacted via email/phone to provide this information.Eligible studies that did not provide IPD data in a way that enabled us to analyse the relationship between intervention, mortality and re-epithelisation time were excluded from the meta-analysis.
The risk of bias was assessed according to the GRADE guidelines 56 to assess the quality of the studies included.Studies were scored according to selection of participants, comparability and outcome and were thus rated from very low to moderate.

Data analysis
A meta-analysis was undertaken examining the association of debridement, dressings, SCORTEN and systemic medications on mortality and re-epithelisation.
A one-step IPD analysis was undertaken.A generalised linear mixed survival model was used to assess the relationship between intervention (dressings vs. debridement) and (a) time to death and (b) time to reepithelialisation.A numerical study identifier was incorporated in the model as a random factor.Sex, year of study and SCORTEN were used as covariates in multivariable models.Age was not adjusted for, as it is used within the calculation for SCORTEN score.Analyses were conducted in R Studio, using the mixed effects COX model (COXME) package.

Data synthesis
A narrative synthesis was employed since the findings from the studies were heterogeneous in terms of design, methods, interventions and outcome measures.The synthesis contains text and tables to summarise the data and explain the treatment outcomes from each intervention.Furthermore, this synthesis covered the main findings of the included studies according to the GRADE framework (Appendix 1). 56

Study selection
This search resulted in 3832 articles (Figure 1): 1563 from SCOPUS, 2247 from PubMed and 22 from Cochrane.Ten additional articles were found through screening the reference lists of the articles resulting in a total of 3832 articles.A total of 1001 studies were removed as duplicates, and 2591 studies were eliminated.186 studies were excluded after screening the full text, leaving 54 studies meeting the inclusion criteria (see Figure 1).In the 54 studies included, we identified and analysed data from a total of 227 individual patients.Dressings for TEN wound management was the most common intervention type (168/227, 74%) compared to 26% (59/227) of studies employing surgical debridement.

Study characteristics
Data were extracted from 227 patients (Table 1).Seventytwo were from Europe, 47 from South America, 46 from Southeast Asia, 42 from the Indian subcontinent, 13 from Australasia, 5 from North America, 2 from Africa.Patient inclusion criteria were reasonably similar across all studies.

Risk of bias within studies
Assessments regarding the risk of bias in the included studies are summarised in the Appendix 1 and rated very low to moderate quality according to GRADE criteria. 56

Results of individual studies
There was no significant association between wound care method and mortality (Table 2, Figure 2).Among people who survived, the median time to re-epithelialisation was faster among people who received dressings (14 days) compared to debridement (17 days), (univariate p = 0.022) (Table 1, Table 3 and Figure 3).
We performed univariate and multivariate analyses assessing time to death and time to re-epithelisation.Our univariate analysis supported the findings of our survival analysis (Figures 2 and 3) where there was no significant difference in mortality between patients who were managed with dressings compared to those receiving debridement (p > 0.05, Table 2).However, those patients who were managed with dressings were found to re-epithelialise almost twice as fast compared to those patients who received debridement (p = 0.023, Table 3).Compared to patients with SCORTEN 1 (reference group), SCORTEN 2 or more was associated with a higher risk of death (p < 0.001, Table 1).
Our univariate analysis found IVIG to be highly protective and significantly reduced mortality by 79% when administered as part of management (HR: 0.21 [0.09-0.45],p < 0.001, Table 2).Cyclosporin was also protective and reduced mortality by 91% when administered   2).However, these therapies did not impact reepithelisation time (p > 0.05, Table 3).The number of patients receiving other therapies (e.g.oral corticosteroids, etanercept) were not sufficient to undertake analysis.
Our multivariate analysis similarly found a mortality benefit from IVIG (p < 0.001, Table 2) and cyclosporin (p = 0.046, Table 2).Again, multivariate analysis did not demonstrate an impact of these medications on reepithelialisation time (p > 0.05, Table 3).

Risk of bias across studies
Bias may have emerged due to the low quality of evidence available and need to incorporate case reports and small cohort studies into our analysis of this understudied area.

DISCUSSION
Despite expert consensus advocating for dressings for wound care in patients with TEN, 23 there is limited high-quality evidence to guide treatment, and there is great variability in approaches across institutions. 20,29,57is systematic review and meta-analysis of IPD aimed to evaluate the optimal approach to wound care in patients with TEN.
Our study found that in patients with TEN, there is faster re-epithelisation in patients who were treated with dressings when compared to surgical debridement, but there was no difference in mortality.This finding is clinically relevant as it suggests dressings accelerate recovery, with implications including a shorter hospital stay, reduced rates of complications and less dependence on resources.This can protect the dermis and promote wound healing. 58e found that patients with 5+ SCORTEN had a mortality rate of 62.5%.This conflicts with the current estimated SCORTEN mortality of >90%.This could be a result of publication bias, or that the most severe TEN patients are palliated instead of referral to a specialised unit for management. 11hile we focussed on wound care, various seminal articles have discussed the efficacy of adjuvant therapy in TEN. 29,59,60Our univariate and multivariate analysis found adjuvant therapies IVIG and cyclosporin were highly protective and significantly reduced mortality regardless of the wound care intervention.This supports the literature where IVIG improved patient outcomes and mortality when incorporated early as part of patient management. 34,61,62,63,64,65However, while demonstrating a mortality benefit, these medications did not change the time to re-epithelialisation, suggesting that wound care may be the most important factor determining reepithelialisation time.Additionally, it is acknowledged that a prospective study standardising wound care is required to answer this question, and these results should be interpreted with caution.
We acknowledge there are several limitations of this study.This study has a heterogeneous pool of studies, limiting the conclusions and generalisability of the findings.As an observational study, there are likely biases, including reporting bias, immortal time bias, unmeasured confounders and problems from non-standardised data capture.However, this evidence is the largest data set available for this rare disease and is the best evidence we can generate from the data available.While a prospective study is desirable, realistically, this would take many years to sufficiently recruit enough patients to power an outcome.However, in the future, a global registry with a standardised case report form could allow global data collection, and multi-centre trials to evaluate the effectiveness of interventions would better determine the evidence for the management of this condition.

CONCLUSION AND IMPLICATIONS
In summary, this systematic review and meta-analysis of the existing literature supports the expert consensus of the dermatology hospitalists; patients with TEN should have the necrotic epidermis left in situ and be supported with dressings, which promotes more rapid re-epithelialisation.

T A B L E 1
Descriptive statistics of TENS independent patient data meta-analysis cohort, by intervention and mortality (2000-2021) (No of participants = 227, no of studies = 54).

T A B L E 2
Results of a frailty model to predict time to death (survival analysis incorporating a random effect for study identifier) (no of participants = 227, no of studies = 54).

T A B L E 3
Results of a frailty model to predict time to re-epithelialisation (survival analysis incorporating a random effect for study identifier) (no of participants = 227, no of studies = 54).

Further
prospective studies are needed for patients with TEN.APPENDIX 1: GRADE assessment of risk of bias and Oxford Centre Quality Rating of evidence for the TENS independent patient data meta-analysis cohort (no of participants = 227, no of studies = 54).

Author, date, country Study type Risk of bias Inconsistency Indirectness Imprecision Publication bias GRADE Oxford rating
A P P E N D I X 1 (Continued)A P P E N D I X 1(Continued)