RDW-SD and PCT Are Potential Prognostic Factors for In-hospital Death in Patients With Stevens-Johnson Syndrome/Toxic Epidermal Necrolysis

Abstract

Purpose

Our study aimed to explore potential prognostic factors in Stevens-Johnson syndrome/toxic epidermal necrolysis (SJS/TEN) patients from easily accessible laboratory data and to investigate whether the combination of these indicators with a score for toxic epidermal necrolysis (SCORTEN) can improve the predictive value.

Methods

Data from 85 SJS/TEN patients hospitalized from 2010 to 2021 were retrospectively analyzed. The primary outcome was in-hospital mortality. Univariate analysis was used to screen for laboratory indexes associated with death. Logistic regression was used to analyze significant risk factors for death. The differentiation and calibration of SCORTEN and modified score were assessed using receiver operating characteristic (ROC) curves and Hosmer-Lemeshow goodness-of-fit test. Net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were used to evaluate the incremental prognostic value.

Results

Among the 85 patients (37 males, 48 females) aged 14-88 years, the mortality rate was 11.8% (n = 10). SCORTEN had good discrimination and calibration to predict mortality in this cohort of patients (area under the ROC curve [AUC] of 0.874, 95% confidence interval [CI], 0.758-0.990; Hosmer-Lemeshow goodness-of-fit test P = 0.994). Red cell distribution width-standard deviation index (RDW-SD) > 47.9 fL and procalcitonin (PCT) > 0.67 ng/mL were significant risk factors for death. When adding the 2 factors to SCORTEN, AUC was 0.915 (95% CI, 0.833–0.997), but not statistically different compared to SCORTEN alone (P = 0.091). The NRI was 1.2 (95% CI, 0.672–1.728; P < 0.001) and the IDI was 0.09 (95% CI, 0.011–0.173; P = 0.026), still suggesting that the modified score had better discriminatory and predictive power than SCORTEN alone. The modified score also showed good calibration (Hosmer-Lemeshow goodness-of-fit test, P = 0.915).

Conclusions

SCORTEN is a good predictor of mortality in SJS/TEN patients in southwest China. Combining RDW-SD > 47.9 fL and PCT > 0.67 ng/mL with SCORTEN may enhance the ability to predict prognosis.

INTRODUCTION

Stevens-Johnson syndrome/toxic epidermal necrolysis (SJS/TEN) is a rare and serious cutaneous, mucosal adverse reaction, usually induced by drugs. Its main clinical manifestations are extensive epidermal detachment of necrotic epidermis and erosions of mucous membrane. SJS/TEN has a high mortality rate because it is prone to various infections and multi-organ damage. Currently, supportive care, systemic corticosteroids, high-dose intravenous immunoglobulin (IVIG), plasmapheresis (PE), cyclosporine, and tumor necrosis factor (TNF)-α inhibitors are commonly used treatments, but the effectiveness of these therapies remains controversial, and treatment strategies vary among different countries.¹,² To assess the severity and predict the prognosis of the disease at the early stage, the severity-of-illness score for toxic epidermal necrolysis (SCORTEN) was developed by French researchers in 2000.³ So far, SCORTEN has been the most widely used and highly recognized scoring system to predict mortality in SJS/TEN patients by 7 clinical and biochemical parameters within 24 hours of admission. Apart from its primary role in predicting prognosis, SCORTEN also functions as an internal control in therapeutic studies for immunomodulatory agents and a benchmark for quality of care across different centers worldwide.⁴,⁵ Despite the overall remarkable accuracy displayed by SCORTEN, some studies still cast doubts on its predictive accuracy.⁶ Some data suggested that SCORTEN was reliable in predicting mortality only in TEN patients with high scores.⁷ In particular, with recent continuous improvements in the medical level, immunomodulator therapy and supportive/intensive care may have impacts on the predictive ability of SCORTEN.

Recently, an alternative risk prediction model, named ABCD-10 (age, bicarbonate, cancer, dialysis, 10% body surface area [BSA]), was proposed by Noe et al. ⁸ It used a multi-institutional cohort of patients in the United States, but it had not been widely validated in various populations. Some studies revealed that ABCD-10 was still inferior to the discrimination performance of SCORTEN.⁹,¹⁰ Other factors that may affect the prognosis of SJS/TEN include chronic kidney disease (CKD),¹¹ tuberculosis,¹² septicemia,¹³ complications of respiratory disorder,¹⁴ etc. In addition, the role of new inflammatory markers, such as red cell distribution width to hemoglobin ratio (RDW/Hb),¹⁵ red blood cell distribution width to albumin ratio (RDW/Alb),¹⁶ and neutrophil to lymphocyte ratio,¹⁷ in the prognosis of SJS/TEN is also a research hotspot in recent years. However, due to the rarity of SJS/TEN and the heterogeneity of the studies, clinical risk factors for its mortality remain unclear.

This study aimed to analyze the ability of SCORTEN to predict mortality in SJS/TEN patients in southwest China using 11-year retrospective data and to find risk factors associated with death in SJS/TEN patients from simple and easily available laboratory indicators. We also investigated whether these risk factors could better predict prognosis when combined with the original SCORTEN model.

MATERIALS AND METHODS

Subjects

SJS/TEN patients hospitalized in Sichuan Provincial People’s Hospital with complete medical records from January 2010 to April 2021 were retrospectively analyzed. Patients who were in compliance with the Bastuji-Garin criteria¹⁸ were included in this study. Patients with the following conditions were excluded: 1) autoimmune bullous diseases, staphylococcal scalded skin syndrome, and other bullous skin diseases; 2) other types of drug eruption; 3) with serious uncontrolled systemic diseases (e.g., severe hepatitis, unstable hematologic disorders, severely active systemic lupus erythematosus); 4) history of blood transfusion within 3 months; and 5) pregnancy and lactation. This study was approved by the Ethics Committee of Sichuan Provincial People’s Hospital (No. 2020-472), and informed consent was waived because of the retrospective nature of the investigation and anonymous data analysis. The study was performed in compliance with the principles of the Declaration of Helsinki.

Data collection and definitions

After screening according to inclusion and exclusion criteria, eligible cases were identified from the inpatient electronic medical record system. The information on age, sex, vital signs, causative drugs, interval time between disease onset and admission, laboratory data, the epidermal detachment area, underlying diseases, complications during hospitalization, hospital stay, and treatment modality, etc., were collected independently by the 2 dermatologists. All the laboratory data for SCORTEN were obtained within 24 hours of admission. SCORTEN scores were calculated with reference to the relevant literature.³ Culprit drugs were identified by a comprehensive review of drug history and algorithm of drug causality for epidermal necrolysis (ALDEN) score.¹⁹

The definitions for complications during hospitalization were as follows: liver injury was defined as any increase in liver enzymes at least 2 times above the upper limit of normal. Confirmation of pneumonia was based on typical symptoms and microbiological and chest imaging evidence. Acute renal failure was diagnosed as serum creatinine (Crea) 1.5 times of known baseline or urine output less than 0.5 mL/kg/h for 6 hours. Gastrointestinal bleeding was judged if black or bloody stools with persistent positive results of stool occult blood test occurred, and the influence of food and medication was excluded. Sepsis was diagnosed based on the sepsis-3 definition.²⁰

Statistical analysis

All the statistical analyses were performed using SPSS version 26.0 (IBM, New York, NY, USA) and R version 4.0.1 (R, Foundation for Statistical Computing, Vienna, Austria). Continuous variables conforming to the normal distribution were expressed as mean ± standard deviation, and those conforming to the skewed distribution were expressed as median (range). Patients were divided into surviving and dead groups for univariate analysis, and the independent sample t-test, Fisher's exact or Mann-Whitney U test were used. The laboratory data with a significance level of P < 0.05 in univariate analysis were included as candidates and were converted into binary variables according to the Youden index for the backward stepwise logistic regression. The binary variables screened by logistic regression were added to the SCORTEN to form a modified model. The receiver operating characteristic (ROC) curves were drawn to evaluate the classification and prediction performances of the scoring model. The Hosmer-Lemeshow goodness-of-fit test was used to assess calibration. The improved prognostic value of the modified model was evaluated by comparing the area under the ROC curve (AUC) and using net reclassification improvement (NRI) and integrated discrimination improvement (IDI).

RESULTS

Patient characteristics

A total of 85 SJS/TEN patients, including 37 males and 48 females, with a male-to-female ratio of 1:1.3, aged 14–88 years (52.36 ± 19.31), of which 17 (20.0%) were ≤ 30 years, 26 (30.6%) were 31–55 years, 34 (40.0%) were 56-75 years, and 8 (9.4%) were > 75 years, were enrolled in this retrospective case series study. Of the 85 patients, 61 were diagnosed with SJS, 6 with SJS/TEN overlap, and 18 with TEN. Ten patients (9 with TEN and 1 with SJS/TEN overlap) died during hospitalization, with an in-hospital mortality rate of 11.8%. Fifty patients had background comorbidities including hypertension, diabetes, gout, CKD, chronic obstructive pulmonary disease (COPD), cardiovascular disease, malignant tumor, epilepsy, mood disorder, etc. The demographic characteristics and underlying diseases of the patients are summarized in Table 1.

Table 1. Demographic characteristics and underlying diseases of SJS/TEN patients.

Characteristics		Total (n = 85)	Survivors (n = 75)	Deceased (n = 10)
Age (yr)		52.4 ± 19.3	49.8 ± 18.8	72.0 ± 9.7
	≤ 30	17 (20.0)	17 (22.7)	0
	31–55	26 (30.6)	26 (34.7)	0
	56–75	34 (40.0)	28 (37.3)	6 (60.0)
	> 75	8 (9.4)	4 (5.3)	4 (40.0)
Sex (male)		37 (43.5)	33 (44.0)	4 (40.0)
Diagnosis
	SJS	61 (71.8)	61 (81.3)	0
	SJS-TEN overlap	6 (7.1)	5 (6.7)	1 (10.0)
	TEN	18 (21.2)	9 (12.0)	9 (90.0)
Underlying diseases
	Hypertension	20 (23.5)	18 (24.0)	2 (20.0)
	Diabetes	11 (12.9)	10 (13.3)	1 (10.0)
	Gout	10 (11.8)	9 (12.0)	1 (10.0)
	CKD	9 (10.6)	7 (9.3)	2 (20.0)
	COPD	9 (10.6)	7 (9.3)	2 (20.0)
	Cardiovascular disease	8 (9.4)	6 (8.0)	2 (20.0)
	Active malignancy	5 (5.9)	4 (5.3)	1 (10.0)
	Epilepsy	4 (4.7)	4 (5.3)	0
	Mood disorder	3 (3.5)	3 (4.0)	0

Values are presented as mean ± standard deviation or number (%).

SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease.

Culprit drugs

Seventy-nine patients were induced by drugs. Table 2 shows the specific culprit drugs. The most common causative drugs were anticonvulsants (18/79, 22.8%), antibiotics (13/79, 16.5%), allopurinol (10/79, 12.7%), non-steroid anti-inflammatory drugs (NSAIDs) (9/79, 11.4%) and traditional Chinese medicines (TCM) (6/79, 7.6%). Six patients (6/79, 7.6%) were sensitized by mixed drugs. Culprit drugs could not be identified in 8 patients (10.1%). The incubation period was 13 (1-40) days.

Table 2. Culprit drugs in patients with SJS/TEN.

Culprit drugs		Number of cases
Anticonvulsants		18 (22.8)
	Carbamazepine	13
	Lamotrigine	3
	Oxcarbazepine	2
Antibiotics		13 (16.5)
	Cephems	5
	Penicillins	3
	Quinolones	2
	Metronidazole	2
	Macrolides	1
	Allopurinol	10 (12.7)
NSAIDS		9 (11.4)
	Ibuprofen	4
	Naproxen	2
	Acetaminophen	2
	Meloxicam	1
Traditional Chinese medicine		6 (7.6)
	Panax notoginseng saponins	3
	Xueshuantong	2
	Chinese herbs	1
Other drugs		9 (10.6)
Mixed drugs		6 (7.6)
Unidentified drugs		8 (10.1)

Values are presented as number (%).

SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis; NSAIDS, non-steroid anti-inflammatory drugs.

Treatment strategies and the predictive value of SCORTEN in SJS/TEN patients

Besides supportive treatment and care, all the patients were given systemic corticosteroids (the initial dose was equivalent to 0.83-25.00 mg/kg∙d prednisone), of which 21 patients (24.7%) received steroid pulse therapy for 3–5 days (500–1,000 mg/day of methylprednisolone). Forty-nine patients (57.7%) were treated with IVIG at the same time (400 mg/kg/day for 3–5 days). Another 3 patients received PE while using corticosteroids and IVIG. We assessed the difference between the predicted mortality and actual mortality of the patients according to SCORTEN. Actual mortality was somewhat lower than predicted mortality at every SCORTEN score level except 6 points, but none of the differences were statistically significant (all P > 0.05) (Table 3). The ROC curve showed an AUC of 0.874 with 95% confidence interval (CI) of 0.758–0.990, and the Hosmer-Lemeshow goodness-of-fit test showed P = 0.994, indicating the good predictive ability and calibration of SCORTEN for death in this group of patients.

Table 3. Comparison of actual and predicted mortality at each SCORTEN level.

SCORTEN	Number of patients	Predicted mortality	Actual mortality	P value
0	19	0.23 (1.2)	0 (0.0)	1.000
1	28	1.09 (3.9)	1 (3.6)	1.000
2	15	1.83 (12.2)	1 (6.7)	0.550
3	12	3.89 (32.4)	2 (16.7)	0.356
4	7	4.35 (62.2)	3 (42.9)	0.606
5	3	2.55 (85.0)	2 (66.7)	0.317
6	1	0.95 (95.1)	1 (100.0)	1.000
Total	85	14.89 (17.5)	10 (11.8)	0.279

Values are presented as number (%).

SCORTEN, score for toxic epidermal necrolysis.

Comparison of clinical parameters between survivors and deceased patients

The comparison of clinical parameters between the survivors and deceased cases is summarized in Table 4. Univariate analyses of the clinical characteristics, laboratory indicators within 24 hours of admission, and complications during hospitalization were conducted to compare the differences between the 2 groups of patients and explore the possible factors affecting death. The results showed that the age and SCORTEN scores were significantly higher in the dead patients than in the survivors. Laboratory indicators of red cell distribution width-standard deviation index (RDW-SD), creatinine, aspartate aminotransferase (AST), C-reactive protein (CRP), and procalcitonin (PCT) levels were significantly higher in the deceased patients than in the survivors, while the level of albumin (Alb) was significantly lower (P < 0.05). Patients who were complicated with pneumonia, acute kidney injury, and sepsis during hospitalization were at higher risk of death (P < 0.05).

Table 4. Comparison of clinical parameters between survivors and deceased patients with SJS/TEN.

Variables		Survivors (n = 75)	Deceased (n = 10)	t/Fisher/Z value	P value
Gender(male)		33 (44.0)	4 (40.0)		1.000
Age (yr)		49.75 ± 18.78	72.00 ± 9.66	−3.668	0.000
Interval between disease onset and admission (days)		5.0 (1–30)	5.0 (1–20)	−0.144	0.885
Incubation period (days)		11.0 (1–40)	14.0 (7–35)	−0.909	0.363
Hospitalization (days)		14.0 (5–33)	10.5 (3–27)	−1.490	0.136
Fever on admission (>38℃)		15 (20.0)	2 (20.0)		1.000
SCORTEN		1.0 (0–5)	4.0 (1–6)	−3.937	0.000
Laboratory data within 24 hours of admission
	WBC (×109/L)	7.77 ± 3.17	8.63 ± 4.37	−0.767	0.445
	LYM (×109/L)	1.05 (0.28–4.76)	1.01 (0.21–3.26)	−1.650	0.099
	EOS (×109/L)	0.32 (0.00–2.80)	0.00 (0.00–0.29)	−1.913	0.056
	Hb (g/L)	132.83 ± 18.99	119.30 ± 27.49	1.509	0.162
	RDW-SD (fL)	43.38 ± 8.37	50.87 ± 9.82	−2.604	0.011
	PLT (×109/L)	187.00 (45.00–443.00)	187.50 (67.00–425.00)	−0.164	0.870
	Crea (umol/L)	69.70 (21.80–620.20)	136.20 (33.10–1184.30)	−1.978	0.048
	Alb (g/L)	36.43 ± 4.86	31.28 ± 2.92	3.264	0.002
	ALT (U/L)	38.00 (3.00–528.00)	38.00 (11.00–257.00)	−1.521	0.128
	AST (U/L)	39.00 (3.00–528.00)	72.50 (14.00–572.00)	−2.142	0.032
	FIB (g/L)	3.82 ± 1.50	3.34 ± 1.31	−0.765	0.444
	CRP (mg/L)	23.30 (0.50–166.00)	57.57 (4.26–170.73)	−2.319	0.020
	PCT (ng/mL)	0.12 (0.05–13.44)	2.78 (0.26–95.88)	−4.114	0.000
	IgE (IU/mL)	70.70 (16.70–3090.00)	216.00 (40.50–529.00)	−1.244	0.214
Complications during hospitalization
	Liver injury	16 (21.3)	4 (40.0)		0.236
	Pneumonia	14 (18.7)	6 (60.0)		0.010
	Acute kidney injury	4 (5.3)	4 (40.0)		0.006
	Gastrointestinal bleeding	5 (6.7)	2 (20.0)		0.190
	Sepsis	5 (6.7)	5 (50.0)		0.001

Values are presented as mean ± standard deviation or median (range) or number (%).

SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis; SCORTEN, score for toxic epidermal necrolysis; WBC, white blood cells; LYM, lymphocytes; EOS, eosinophils; Hb, hemoglobin; RDW-SD, red cell distribution width-standard deviation index; PLT, platelet; Crea, creatinine; Alb, albumin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; FIB, fibrogen; CRP, C-reactive protein; PCT, procalcitonin; IgE, immunoglobulin E.

Possible refinement of SCORTEN

Because the purpose of this study was to find risk factors for death in SJS/TEN patients among easily accessible laboratory indices, we only considered laboratory findings within 24 hours of admission that were statistically significant in the univariate analysis as candidates. Although some of the complications that occurred during hospitalization like pneumonia, acute kidney injury, and sepsis also suggested a higher risk of death, we did not consider these complications in this cohort of patients suitable as possible improvement factors for SCORTEN because they usually occurred several days after admission. In contrast, the components in SCORTEN were collected within 24 hours of admission. All variables with statistical significance in Table 4 were dichotomized according to the Youden index (RDW-SD > 47.9 fL, Crea > 113.5 umol/L, Alb < 35.1 g/L, AST > 43.0 U/L, CRP > 30.3 mg/L, PCT > 0.67 ng/mL), and then the 6 variables were included as candidates in the multivariable model. Based on backward stepwise logistic regression, RDW-SD > 47.9 fL and PCT > 0.67 ng/mL were significant risk factors for in-hospital death (Table 5). Combining RDW-SD > 47.9 fL and PCT > 0.67 ng/mL with SCORTEN increased AUC from 0.874 (95% CI, 0.758–0.990) to 0.915 (95% CI, 0.833–0.997) in predicting in-hospital mortality, but there was no significant difference (P = 0.09) (Figure). The NRI and IDI were further applied to compare the discriminative and predictive power of the 2 models. The results showed that continuous NRI was 1.2 (95% CI, 0.672-1.728, P < 0.001) and IDI was 0.09 (95% CI, 0.011–0.173, P = 0.026), indicating that adding RDW-SD > 47.9 fL and PCT > 0.67 ng/mL acquired greater discriminative and predictive ability compared with using SCORTEN alone. The new model also had a similar calibration ability to the original SCORTEN model (Hosmer-Lemeshow goodness-of-fit test, P = 0.915 and P = 0.994, respectively).

Table 5. Multivariable analysis based on backward stepwise logistic regression for in-hospital mortality in patients with SJS/TEN.

Risk factors	OR (95% CI)	P value
RDW-SD > 47.9 fL	6.21 (1.027, 37.539)	0.047
PCT > 0.67 ng/mL	19.64 (2.063, 187.923)	0.010

SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis; OR, odds ratio; CI, confidence interval; RDW-SD, red cell distribution width-standard deviation index; PCT, procalcitonin.

Figure. Sensitivity and specificity of SCORTEN and the modified score in predicting mortality for patients with SJS/TEN.

SCORTEN, score for toxic epidermal necrolysis; SJS, Stevens-Johnson syndrome; TEN, toxic epidermal necrolysis.

DISCUSSION

SJS/TEN is a rare immune-mediated severe cutaneous/mucosal reaction characterized by epidermal detachment due to apoptosis and necrosis of keratinocytes, with SJS involving < 10% of the body surface area (BSA) exfoliation, SJS/TEN overlap involving 10%–30% of the BSA, and TEN involving > 30% of the BSA. The rarity of SJS/TEN makes it more difficult to conduct a large randomized controlled and comparative trial. Hence, most current clinical knowledge and management guidelines about SJS/TEN are based on retrospective studies or case series. There is a certain gender difference in SJS/TEN patients. Most previous studies showed a female predominance,¹⁴,²¹ consistent with the present study, but there were still a few literatures to show that males are more susceptible than females.²² Some researchers found that female patients had worse prognosis and higher mortality,²³ but we did not find a correlation between gender and death. SJS/TEN has a wide age distribution, ranging from young children to the elderly. It is reported that patients older than 70 years have the highest risk.²⁴ In our study, the highest number of patients was in the age of 56-75 years, which to some extent confirmed that the elderly had a higher risk to develop SJS/TEN. Drugs are responsible for the vast majority of SJS/TEN cases, and this cohort was no exception. The most common causative drugs in this study are anticonvulsants, antibiotics, allopurinol, and NSAIDs, which are similar to the results reported in previous literature.²⁵,²⁶ In addition, it should be noted that SJS/TEN induced by TCM also accounts for a certain proportion. In some Asian countries, TCM is highly accepted by the people and is generally considered to have low toxic and side effects, while there have recently been more and more reports of adverse reactions to TCM at home and abroad, so TCM should be used with special caution.

SJS/TEN is a serious and rapidly progressive disease, leading to organ dysfunction and high mortality. The mortality rate of SJS/TEN reported in Germany was as high as 34%,²⁷ while that of SJS and TEN reported in Japan was 4.1% and 29.9% respectively.²⁸ To date, the treatment of SJS/TEN still poses a great challenge. There is no consensus on the treatment strategy for SJS/TEN, and commonly used drugs include glucocorticoids, IVIG, cyclosporine, and TNF-α inhibitors. Although the therapeutic efficacy of glucocorticoids is controversial, several studies have suggested that it is still the main therapeutic agents for SJS/TEN. Short-term application of medium- to high-dose of systemic applications of glucocorticoids can effectively treat SJS/TEN with mild complications and little effect on underlying diseases.²⁹ It has been reported that corticosteroids and IVIG combination therapy has significant survival benefits for patients with SJS/TEN.³⁰,³¹ In our study, all patients were treated with glucocorticoids and more than half combined with IVIG; steroid pulse therapy was administered in severe and rapidly progressive cases. Although there was no statistical significance, the mortality of the 85 patients had a lower trend compared to that predicted by SCORTEN, indicating that glucocorticoid therapy or combined with IVIG may help improve patient prognosis to some extent. Prospective randomized clinical trial is the best way to validate this, but it is very difficult to conduct because of the rarity of SJS/TEN.

The scoring system called SCORTEN, proposed in 2000, is currently accepted as the disease severity score, and its excellent mortality prediction performance has been validated in studies in different countries.³²,³³ However, the accuracy of SCORTEN may be partially compromised as the level of medical care and diagnostic capabilities continue to improve. A recent study has shown a significantly lower number of deaths than those expected according to SCORTEN.³⁴ Therefore, it has been suggested that the current scoring system may need to be revised or improved when predicting mortality in SJS/TEN patients. Some attempts have been made in recent years to create modified score models with better predictive performance by adding some new indicators to original SCORTEN (e.g., the interval between disease onset and treatment initiation at the specialty hospital (≥ 8 days) and respiratory disorder within 48 hours after admission¹⁴ or RDW/Hb¹⁵), but none of these new models have been well validated. The aim of our study was to obtain possible mortality risk factors among simple laboratory indicators and whether these factors could improve the SCORTEN model. These indicators are simple and easily available, do not require to be calculated, and may be more convenient for clinical use. Patients with SJS/TEN usually present with internal organ damage and systemic bacterial infections, such as impaired renal function, pulmonary infections, liver injury, and hypoalbuminemia, which may indicate poor prognosis.¹¹,³⁵,³⁶,³⁷,³⁸ The results of this study showed that RDW-SD, Crea, AST, CRP, PCT were significantly higher and Alb was significantly lower in deceased patients compared to survivors. These indicators reflect the status of liver and kidney function, nutritional status, systemic bacterial infections, and inflammatory reactions to a certain extent. The differential performance of these factors in surviving and deceased patients also conforms to the actual situation that SJS/TEN patients are prone to multi-organ involvement and systemic infections. Further multivariable regression analysis revealed that RDW-SD > 47.9 fL and PCT > 0.67 ng/mL were significant risk factors for death during hospitalization based on backward stepwise logistic regression. A previous study of our research group has shown that PCT is a valuable index in detecting systemic bacterial infections in SJS/TEN patients, and its level can partially reflect the severity of the disease.³⁹ In addition, elevated PCT and hypothermia were found to be useful markers for the timely detection of bacteremia of SJS/TEN patients in another report.⁴⁰ Some researchers suggest that day 0 PCT levels should be considered as an independent prognostic marker for SJS/TEN.⁴¹ A recent study found that RDW/Hb showed similar predictive accuracy with SCORTEN in patients with SJS/TEN, and when used in combination with SCORTEN, it also contributed to enhanced prognostic ability.¹⁵ Another study revealed that RDW/Alb could be a meaningful biomarker to assess the severity of SJS/TEN.¹⁶ However, there have been no studies showing the role of RDW alone in the prognosis of SJS/TEN. In this study, RDW-SD47.9 fL and PCT > 0.67 ng/mL were screened out as significant risk factors for in-hospital death from the meaningful indicators of univariate analysis, which can partly corroborate these findings in the appeal. The 2 factors were further added to SCORTEN and the new addition was found to improve the AUC of SCORTEN for predicting mortality and showed better model discrimination by calculating NRI and IDI, suggesting that RDW-SD and PCT can be considered as possible new indicators for improving SCORTEN, but this requires further validation on a larger scale and in different populations. The precise mechanism in which RDW-SD and PCT act as prognostic factors for SJS/TEN is unclear. We speculated that in a status of severe systemic inflammatory response in SJS/TEN, erythrocyte homeostasis was probably altered and ineffective erythropoiesis was mediated¹⁵,⁴²; the synthesis of PCT might induced by pro-inflammatory cytokines such as TNF-α and IL-6.⁴³ Of course, these speculations need to be confirmed by further studies.

Furthermore, patients who developed pneumonia, acute kidney injury, and sepsis during hospitalization had a higher risk of death in this cohort of patients, suggesting that the presence of systemic infection and renal dysfunction during treatment were important predictive factors for poor prognosis, which is consistent with previous studies.⁴⁴,⁴⁵ Multi-organ failure, generalized infections and sepsis were also confirmed to be the major causes of mortality in SJS/TEN patients.⁴⁶,⁴⁷,⁴⁸ These complications may be related to untimely control of the disease, the use of high-dose corticosteroids, secondary infections, hypovolemia and so on. Since they occurred several days or more after admission, and the components of SCORTEN were collected just on admission, we did not include them as candidates for refining SCORTEN. However, these results suggested that in SJS/TEN patients, in addition to general care and immunomodulatory therapy, close monitoring and attention should be paid to skin and systemic infections, loss of body fluid, electrolyte imbalance, thermoregulatory dysfunction, nutritional support, stabilization and maintenance of visceral organ function. Early identification and appropriate management of these possible complications are important.

The study has inevitable limitations. First, it was a single-center and retrospective study with a relatively small sample size. Secondly, data were only collected during the first 24 hours after admission and SCORTEN scores on different times such as on days 3 and 5 were not assessed, because of lack of these data. Furthermore, the predictive value of the modified model needs to be validated in different cohorts. Although it is difficult to implement, multi-center prospective and large sample size studies would go a long way in addressing the current limitations.

In conclusion, our study showed that SJS/TEN patients were predominantly female and the largest number of patients were between 56 and 75 years of age, with an in-hospital mortality rate of 11.76%. The most common causative drugs were anticonvulsants, antibiotics, and allopurinol. SCORTEN is still a good predictor of mortality in SJS/TEN patients in southwest China. RDW-SD and PCT may be potential indicators to improve the predictive value of SCORTEN and adding RDW-SD > 47.9 fL and PCT > 0.67 ng/mL to the original SCORTEN could elevate its discriminatory and predictive power.