If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Taipei, Linkou and Keelung, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan
Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Taipei, Linkou and Keelung, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan
Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Taipei, Linkou and Keelung, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan
Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Taipei, Linkou and Keelung, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan
Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Taipei, Linkou and Keelung, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan
Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Taipei, Linkou and Keelung, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan
Department of Laboratory Medicine, Chang-Gung Memorial Hospital, Taoyuan County, TaiwanDepartment of Medical Biotechnology and Laboratory Science, Chang Gung University, Taoyuan County, Taiwan
Department of Laboratory Medicine, Chang-Gung Memorial Hospital, Taoyuan County, TaiwanDepartment of Medical Biotechnology and Laboratory Science, Chang Gung University, Taoyuan County, Taiwan
Department of Laboratory Medicine, Chang-Gung Memorial Hospital, Taoyuan County, TaiwanDepartment of Medical Biotechnology and Laboratory Science, Chang Gung University, Taoyuan County, Taiwan
Corresponding author at: Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Keelung, Linko, Chang Gung University College of Medicine, 199, Tung-Hwa North Road, Taipei, Taiwan. Tel.: +886 2 27135211x3397, 3399; fax: +886 2 27191623.
Department of Dermatology, Drug Hypersensitivity Clinical and Research Center, Chang Gung Memorial Hospitals, Taipei, Linkou and Keelung, TaiwanCollege of Medicine, Chang Gung University, Taoyuan, Taiwan
Increasing studies reported genetic susceptibility to drug hypersensitivity reactions, as exemplified by the HLA-A*31:01 and HLA-B*15:02 association with carbamazepine (CBZ)-induced hypersensitivity reactions, such as maculopapular exanthema (MPE), drug rash with eosinophilia and systemic symptoms (DRESS), and Stevens–Johnson syndrome (SJS)/toxic epidermal necrolysis (TEN).
Objective
To carry out a comprehensive analysis on the clinical spectrum and HLA genotype–phenotype correlations in CBZ-induced hypersensitivity reactions.
Methods
We analyzed the clinical information of 194 patients with CBZ hypersensitivity (51 MPE, 23 DRESS, 112 SJS/TEN, and 8 cases with other phenotypes), and 152 CBZ-tolerant controls. All are Han Chinese. We examined the HLA-A/HLA-B genotypes, gene dosage, and drug dosage effects.
Results
CBZ-SJS/TEN showed the strongest association with the HLA-B*15:02 allele (Pc = 5.8 × 10−43; odds ratio (OR) (95% CI) = 97.6(42.0–226.8)), in which HLA-B*15:02 was identified in all patients (25/25) with SJS/TEN with >5% body surface area (BSA) skin detachment, but lost its 100% association (85.1%, 74/87) in SJS with <5% BSA detachment. In contrast, HLA-B*40:01 showed negative association with CBZ-induced SJS/TEN ((Pc = 8.3 × 10−5; OR (95% CI) = 0.22(0.1–0.4)). By comparison, CBZ-induced MPE/DRESS had no association with HLA-B*15:02, but linked to HLA-A*31:01 (Pc = 2.7 × 10−3; OR (95% CI) = 6.86(2.4–19.9), and HLA-B*51:01 (Pc = 0.01; OR (95% CI) = 4.56(2.0–10.5)). No gene dosage or CBZ dosage effects was observed.
Conclusion
This study reported the different strength of HLA association with CBZ hypersensitivity in Han Chinese. With the increasing application of pharmacogenetic markers, the HLA genotype–phenotype correlations and the results of the test need to be carefully interpreted for CBZ-induced hypersensitivity reactions.
]. The clinical presentations of CBZ hypersensitivity range from mild skin rash, such as maculopapular exanthema (MPE), to potentially life-threatening reactions, including drug rash with eosinophilia and systemic symptoms (DRESS, also known as drug induced hypersensitivity syndrome (DIHS)), Stevens–Johnson syndrome (SJS), and toxic epidermal necrolysis (TEN) [
]. Such a strong association has been validated in many populations, especially in the countries of Southeast Asia, such as Thailand, Malaysia, Singapore, Hong Kong, and India, etc. [
Association of HLA-B*1502 allele with carbamazepine-induced toxic epidermal necrolysis and Stevens–Johnson syndrome in the multi-ethnic Malaysian population.
]. The Food and Drug Administration (FDA) of US and similar regulatory agencies of other countries relabeled the drug information of CBZ, and advised physicians to screen HLA-B*15:02 before starting the treatment of CBZ in patients with Asian ancestry [
]. In Taiwan, the cost for HLA-B*15:02 genetic test for the new users of CBZ has been covered by the National Health Insurance since 2010, and the cases of CBZ-SJS/TEN have dramatically decreased thereafter [
]. More recently, HLA-A*31:01 was reported to be associated with CBZ–induced hypersensitivity reactions in multiple ethnicities, including Japanese, Europeans, and Koreans [
Genome-wide association study identifies HLA-A*3101 allele as a genetic risk factor for carbamazepine-induced cutaneous adverse drug reactions in Japanese population.
With the increasing application of pharmacogenetic tests for preventing adverse reactions, a comprehensive analysis on the HLA genotype–phenotype correlations in the clinical spectrum of CBZ-induced hypersensitivity reactions, however, is lacking. The many previous studies on CBZ hypersensitivity reactions were limited by the low incidences, difficulty of patients enrollment, and small sample sizes, making it difficult to investigate the genotype–phenotype correlations [
Association of HLA-B*1502 allele with carbamazepine-induced toxic epidermal necrolysis and Stevens–Johnson syndrome in the multi-ethnic Malaysian population.
Genome-wide association study identifies HLA-A*3101 allele as a genetic risk factor for carbamazepine-induced cutaneous adverse drug reactions in Japanese population.
]. In this study, we carried out a detailed analysis on the clinical information of a total of 194 cases with CBZ-induced hypersensitivity reactions, and investigated the strength of HLA association, clinical correlations, and gene or drug dosage effects on the disease severity.
2. Materials and methods
2.1 Subjects
We collected the clinical data and HLA-A/HLA-B genotype data of a total of 194 patients with a diagnosis of CBZ-induced hypersensitivity reactions from Chang Gung Memorial Hospitals, Wan Fang Hospital, Taipei Medical University, and Chung Shan Hospital in Taiwan. Of the 194 cases, 87 cases (including 56 SJS/TEN, 13 DRESS, and 18 MPE), were previously reported [
]. In addition, we included the data of 107 cases of CBZ-induced hypersensitivity reactions, including 56 SJS/TEN, 10 DRESS, 33 MPE, and 8 patients with other presentations (fixed drug eruption (FDE), erythema multiforme (EM) or pseudolymphoma). In the previous cohort [
], most patients met the definitive SJS/TEN criteria and had more extensive skin detachment. By comparison, in the follow-up enrollment, we included the data of probable SJS patients with limited skin detachment, and there were 85.7% (48/56) patients with <5% BSA skin detachment. We also collected the clinical data and HLA-A, HLA-B genotype data of 152 CBZ-tolerant individuals as the control group, who had received CBZ for at least 3 months without evidence of adverse reactions and been recruited from the same hospital system from Taiwan. Of the 152 CBZ-tolerant controls, 144 were previously reported [
]. All of the subjects were Han Chinese in Taiwan. This study was approved by the institutional review board (IRB) from the ethical standards committee of each studying site/institute before initiation, and was in compliance with the Declaration of Helsinki as well as any relevant local laws, regulations, and guidelines for the use of human subjects.
2.2 Case assessment and clinical information
All cases were assessed by at least two dermatologists who evaluated the clinical presentations, histological data (if available) and clinical information. The drug causality was determined by the Naranjo algorithm and the algorithm of drug causality assessment for SJS/TEN (ALDEN) [
ALDEN, an algorithm for assessment of drug causality in Stevens–Johnson Syndrome and toxic epidermal necrolysis: comparison with case-control analysis.
]. Briefly, these assessment methods included prior drug reaction history, clinical manifestations of typical drug reactions, chronology or temporal relationship between drug use and onset of reaction, re-challenge, de-challenge or improvement after discontinuation of suspected drugs, and the notoriety of suspected drugs. Only cases with probable to definite causes of CBZ (ALDEN score ≥ 4 and Naranjo algorithm ≥ 5) were recruited into this study.
Phenotypes were classified as defined by the RegiSCAR study group [
Correlations between clinical patterns and causes of erythema multiforme majus, Stevens–Johnson syndrome, and toxic epidermal necrolysis: results of an international prospective study.
]. SJS and TEN (SJS/TEN) are characterized by a rapidly developing blistering exanthema of purpuric macules and target-like lesions accompanied by mucosal involvement and skin detachment, in which SJS has less than 10% body surface area (BSA) skin detachment, TEN with more than 30%, and SJS-TEN overlap with 10–29% [
Correlations between clinical patterns and causes of erythema multiforme majus, Stevens–Johnson syndrome, and toxic epidermal necrolysis: results of an international prospective study.
]. The criteria and scoring system of DRESS include the follows: cutaneous involvement with typical skin rash (e.g., exfoliative dermatitis, diffuse maculopapular exanthema), fever, eosinophilia, lymph node enlargement, atypical lymphocytes, internal organ involvement (liver, kidney, central nervous system, lung heart, muscle), time of resolution, and the evaluation of other potential causes [
]. Only the probable or definite cases of DRESS were enrolled in this study.
The MPE phenotype is characterized by generalized cutaneous erythematous macules and papules and is self-limited without systemic involvement. We also enrolled patients with other presentations, including FDE, pseudolymphoma, mucosal EM, atypical EM or atypical SJS induced by CBZ. FDE refers to recurrent cutaneous eruptions at the previously identical affected sites with the subsequent causative agents re-exposure and residual hyperpigmentation after healing [
]. Drug-induced pseudolymphoma is diagnosed with variable cutaneous lesions with histologically atypical lymphocytic infiltration mimicking lymphoma. However, the lesion regresses following the withdrawal of the causative medication(s) [
]. Mucosa EM (also known as Fuchs syndrome) is defined as painful erosions over oral, genital, or ocular mucosa without skin lesions, excluding autoimmune bullous disease and inflammatory dermatosis [
]. Atypical EM includes typical targets or raised edematous papules that are predominantly distributed at non-acral lesions with involvement of one or more mucous membranes. Atypical SJS applies to patients who have developed the typical mucous membrane changes seen in SJS but lacked typical cutaneous lesions.
2.3 HLA-A and HLA-B genotyping
HLA-A and HLA-B genotypes were determined by sequence-specific oligonucleotide reverse line blots (DYNAL Biotech Ltd, Bromborough, UK) or SeCore® HLA Sequence based typing (Invitrogen, Life technologies, USA). Potential ambiguities were resolved by sequence-based typing [
Comparisons of allele frequencies between groups were performed using Fisher's exact tests. All P-values were two-tailed. The corrected P (Pc) values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A and 36 for HLA-B). Pc < 0.05 was considered to be statistically significant. Odds ratios (ORs) were calculated using Haldane's modification, which added 0.5 to all cells to accommodate possible zero counts. Positive post-test probability was calculated using the positive likelihood ratio, and the negative post-test probability was calculated using the negative likelihood ratio. Graphs of conditional probabilities relating pre- and post-test probabilities were generated from likelihood ratios according to the methods previously described [
3.1 Clinical characteristics of the study subjects
We analyzed the clinical information and HLA-A and HLA-B genotyping data of a total of 194 patients with CBZ-induced hypersensitivity reactions, including 112 SJS/TEN, 23 DRESS, 51 MPE, and 8 patients with others presentations (Table 1 and eTable 1). We compared the data of the 194 patients with that of 152 CBZ-tolerant controls. Of the 194 patients, 112 (57.7%) were classified as SJS/TEN, including 96 (49.5%) as SJS, 10 (5.1%) as SJS/TEN overlap, and 6 (3.1%) as TEN; 51 (26.3%) were defined as MPE, 23 (11.9%) as DRESS and 8 (4.1%) as other presentations, including 3 with fixed drug eruption (FDE), 2 with pseudolymphoma, 1 with mucosal erythema multiforme (EM), 1 with atypical EM and 1 with atypical SJS (Table 1 and eTable 1). As 86.5% (96/112) of our SJS/TEN patients had skin detachment less than 10% total BSA, the SJS group was further divided into 3 groups according to area of skin detachment (≤1%, 2–5%, 6–9% of BSA) (Table 1). The most common type in our SJS/TEN group was SJS with 2–5% BSA skin detachment (n = 57, 29.4%), followed by SJS with ≤1% BSA (n = 30, 15.5%) (Table 1). The mean age of patients with CBZ- induced hypersensitivity was 49.2 years-old (range: 7–88) without specific gender distribution (male:female = 102:92) (Table 1). Representative photographs of TEN, SJS with different ranges of BSA skin detachment, mucosal EM, FDE, pseudolymphoma, and DRESS are shown in eFig. 1.
Table 1Characteristics of 194 patients with carbamazepine–induced hypersensitivity reactions.
Abbreviations: DRESS, drug rash with eosinophilia and systemic symptoms; MPE, maculopapular exanthema; other presentations: fixed drug eruption, erythema multiforme, pseudolymphoma, or atypical SJS; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
Abbreviations: DRESS, drug rash with eosinophilia and systemic symptoms; MPE, maculopapular exanthema; other presentations: fixed drug eruption, erythema multiforme, pseudolymphoma, or atypical SJS; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
Abbreviations: DRESS, drug rash with eosinophilia and systemic symptoms; MPE, maculopapular exanthema; other presentations: fixed drug eruption, erythema multiforme, pseudolymphoma, or atypical SJS; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
Abbreviations: DRESS, drug rash with eosinophilia and systemic symptoms; MPE, maculopapular exanthema; other presentations: fixed drug eruption, erythema multiforme, pseudolymphoma, or atypical SJS; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
Abbreviations: DRESS, drug rash with eosinophilia and systemic symptoms; MPE, maculopapular exanthema; other presentations: fixed drug eruption, erythema multiforme, pseudolymphoma, or atypical SJS; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
a Abbreviations: DRESS, drug rash with eosinophilia and systemic symptoms; MPE, maculopapular exanthema; other presentations: fixed drug eruption, erythema multiforme, pseudolymphoma, or atypical SJS; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
3.2 CBZ-induced SJS/TEN showed the strongest association with the HLA-B*15:02 allele
Comparing the difference of HLA-B allele frequencies between the 194 patients with CBZ-induced hypersensitivity and 152 CBZ-tolerant controls, HLA-B*15:02 showed the most significant association with CBZ-induced SJS/TEN (P = 1.6 × 10−44; Pc = 5.8 × 10−43; OR (95% CI) = 97.6 (42.0–226.8)) (Table 2). In particular, the HLA-B*15:02 allele was identified in all patients (25/25) of CBZ-SJS/TEN with >5% BSA skin detachment, but lost its 100% association in SJS with less BSA involvement: 85.1% (74/87) in SJS patients with ≤5% BSA detachment, 87.7% (50/57) in SJS with 2–5% BSA detachment, and 80% (24/30) in SJS with ≤1% BSA detachment (Fig. 1). Two cases among the list of other presentations (25% (2/8), one FDE and one mucosal EM) were also positive for HLA-B*15:02 (Fig. 1 and eFig. 2). However, only 7.8% (4/51) of MPE, 0% (0/23) of DRESS, and 7.2% (11/152) of the CBZ-tolerant controls carried the HLA-B*15:02 allele (Fig. 1). The allele frequency of HLA-B*15:02 observed in CBZ-tolerant controls is similar to that of our general population [
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
55:02
(B55)
9
1
0.048
NS
0.14(0.02–1.1)
1
1.00
NS
0.72(0.1–6.0)
4
0.74
NS
1.35(0.4–4.6)
5
0.77
NS
1.15(0.4–3.6)
Abbreviations: CBZ carbamazepine; DRESS drug rash with eosinophilia and systemic symptoms; MPE maculopapular exanthema; NS no significance; SJS/TEN Stevens–Johnson syndrome/toxic epidermal necrolysis.
* The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (36 for HLA-B). Bold text indicates an unadjusted P-value of less than 0.05.
Fig. 1Analysis of the association between HLA-B*15:02 allele and different types of carbamazepine CBZ)-induced hypersensitivity reactions against tolerant controls. There are four types of CBZ-induced hypersensitivity reactions: SJS/TEN, MPE, DRESS, and other presentations. SJS was further divided into 3 groups according to area of skin detachment (≤1%, 2–5%, 6–9% of BSA). In the column of odds ratio, the values indicate the odds ratios, and the horizontal lines indicate 95% confidence intervals. BSA, body surface area; CI, confidence interval; DRESS, drug rash with eosinophilia and systemic symptoms; HLA, human leukocyte antigen; MPE, maculopapular exanthema; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
A sub-group analysis of the strength of HLA-B*15:02 association with different types of CBZ-induced hypersensitivity reactions showed higher odds ratios (OR) in the groups of CBZ-induced TEN (OR(95%CI) = 160.0 (8.5–3021.5), SJS-TEN overlap (258.4 (14.2–4696.3)), and SJS patients with 6–9% BSA detachment (233.8 (12.8–4277.3)). However, the OR of HLA-B*15:02 association decreased with the less involvement of skin detachment in the SJS patients: 2–5% BSA detachment (OR(95%CI) = 91.6 (33.6–249.1), and ≤1% BSA detachment (51.3 (17.3–151.7) (Fig. 1). HLA-B*15:02 showed no association with MPE, DRESS, or other presentations (Fig. 1).
3.3 The probabilities of HLA-B*15:02 genetic test for various phenotypes of CBZ-induced hypersensitivity reactions
We assessed the probabilities of HLA-B*15:02 genetic test for various CBZ-induced hypersensitivity reactions, and generated graphs shown in Fig. 2. Briefly, we calculated the specificity, sensitivity, likelihood ratios, positive/negative predictive values, and the post-test probability of disease in any patient with a positive or negative test result in each sub-phenotype of CBZ-induced hypersensitivity reactions (Table 3). The HLA-B*15:02 allele has 100% high sensitivity for testing CBZ-induced SJS/TEN in subjects with more than 10% BSA involvement of skin detachment, and decreases the sensitivity to 87% in patients with <10% BSA involvement (Table 3). As a test for CBZ-induced SJS/TEN, HLA-B*15:02 allele has high probabilities, especially with the highest for SJS/TEN with ≥10% BSA of skin detachment, in which a post-test probability for a positive HLA-B*15:02 test is 3.33%, and 0% for the negative result (positive likelihood ratio (LR+) = 13.81 (95% CI:7.53–23.82); negative likelihood ratio (LR−) = 0 (95% CI:0.002–0.5)) (Table 3, Fig. 2). By comparison, HLA-B*15:02 allele shows very low probabilities for testing CBZ-induced DRESS or CBZ-MPE (Table 3, Fig. 2).
Fig. 2The pre- and post-test probabilities of HLA-B*15:02 screening test for various carbamazepine (CBZ)-induced hypersensitivity reactions. Data of five groups were presented, including SJS/TEN (all), SJS/TEN with >10% BSA skin detachment, SJS with <10% BSA skin detachment, MPE and DRESS. Post-test probability was estimated by likelihood ratio. The blue lines represents positive test results (positive likelihood ratio (LR+)), and the red line represents negative likelihood ratio (LR−). When HLA-B*15:02 was tested positive, post-test probability was high for all SJS/TEN. When HLA-B*15:02 was negative, post-test probability was lowest for SJS/TEN with >10% BSA skin detachment. Such relationship was not observed in MPE/DRESS cases. DRESS, drug rash with eosinophilia and systemic symptoms; MPE, maculopapular exanthema; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
Table 3Assessment of HLA-B*15:02 genetic test accuracy in different types of carbamazepine-induced hypersensitivity reactions.
HLA-B*15:02
Sensitivity
Specificity
PPV
NPV
LR+
LR−
Post-test probability for a positive test
Post-test probability for a negative test
SJS/TEN (all spectrum)
0.88
0.93
0.9
0.92
12.21
0.13
2.96%
0.03%
SJS/TEN (≥10%)
1.0
0.93
0.59
1.0
13.81
0
3.33%
0%
SJS (<10%)
0.87
0.93
0.88
0.92
11.95
0.15
2.89%
0.04%
MPE
0.08
0.93
0.27
0.75
1.08
0.99
NA
NA
DRESS
0
0.93
0
0.86
0
1.08
NA
NA
Abbreviations: DRESS drug rash with eosinophilia and systemic symptoms; LR+ positive likelihood ratio; LR− negative likelihood ratio; MPE maculopapular exanthema; NA not available due to no information on the incidence rate; NPV negative predictive value; PPV positive predictive value; SJS/TEN Stevens–Johnson syndrome/toxic epidermal necrolysis.
], a positive result of HLA-B*15:02 test would increase the post-test probability of CBZ-SJS/TEN to 2.96% on the basis of the positive likelihood ratio (LR+) of 12.21 (95% CI: 7.46–20.66). Since the negative likelihood ratio (LR−) was 0.13 (95% CI: 0.08–0.19), a negative result of HLA-B*15:02 test would reduce the probability of CBZ-SJS/TEN from 0.25% to 0.03% (Table 3).
3.4 Association between other HLA-B alleles and various types of CBZ-induced hypersensitivity reactions
We also analyzed the association between other HLA-B alleles and CBZ-induced hypersensitivity reactions. HLA-B*40:01 had a negative association with CBZ-induced SJS/TEN (P = 2.3 × 10−6; Pc = 8.3 × 10−5;OR (95% CI) = 0.22(0.1–0.4) (Table 2). By comparison, HLA-B*51:01 showed significant association with CBZ-induced MPE (P = 0.0006; Pc = 0.02; OR (95% CI) = 4.86(2.0–11.9), as well as CBZ-induced MPE/DRESS (P = 0.0004; Pc = 0.01; OR (95%CI) = 4.56(2.0–10.5))(Table 2).
In addition, we performed sub-group analysis for the CBZ-induced SJS/TEN patients who did not carry HLA-B*15:02. We removed the HLA-B genotype data of 99 CBZ-SJS/TEN patients from the dataset, and analyzed the HLA-B association between 13 patients with CBZ-SJS and the 141 CBZ-tolerant controls (eTable 2). We found that HLA-B*15:11, which belongs to HLA-B75, the same serotype as that of HLA-B*15:02, showed a weak association with CBZ-induced SJS (P = 0.007), yet no significance after correction by multiple comparison (Pc = 0.21; OR(95% CI) = 61.52(2.8–1359.3) (eTable 2).
3.5 Association between HLA-A alleles and various phenotypes of CBZ-induced hypersensitivity reactions
We compared the difference of HLA-A allele frequencies between the 194 patients with CBZ-hypersensitivity and 152 CBZ-tolerant controls. HLA-A*31:01 showed significant association with CBZ-induced DRESS (P = 1.4 × 10−4; Pc = 2.4 × 10−3; OR (95% CI) = 12.9 (3.7–45.2)), as well as CBZ-induced MPE/DRESS (P = 1.6 × 10−4; Pc = 2.7 × 10−3; OR (95% CI) = 6.86 (2.4–19.9)) (Table 4, Fig. 3). By comparison, only 1.8% (2/112) of CBZ-SJS/TEN and 3.3% (5/152) of the CBZ-tolerant control carried the HLA-A*31:01 allele (Table 4, Fig. 3). None of HLA-A alleles had association with CBZ-SJS/TEN (Table 4, Fig. 3). Two cases among the list of other presentations (25% (2/8), one pseudolymphoma and one atypical SJS) were also positive for HLA-A*31:01 (Fig. 3).
Table 4Association between HLA-A alleles and carbamazepine (CBZ)-induced hypersensitivity reactions.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
33:03
(A33)
30
15
0.18
NS
0.63(0.32–1.2)
4
1.00
NS
0.86 (0.27–2.7)
8
0.55
NS
0.76(0.32–1.7)
12
0.77
NS
0.59(0.38–1.6)
Abbreviations: CBZ carbamazepine; DRESS drug rash with eosinophilia and systemic symptoms; MPE maculopapular exanthema; NS no significance; SJS/TEN Stevens–Johnson syndrome/toxic epidermal necrolysis.
* The association was examined by two-tailed Fisher's exact test and the Pc values were adjusted by using Bonferroni's correction for multiple comparisons (17 for HLA-A). Bold text indicates an unadjusted P-value of less than 0.05.
Fig. 3Analysis of the association between HLA-A*31:01 allele and different types of carbamazepine-induced hypersensitivity reactions against tolerant controls. There are four types of CBZ-induced hypersensitivity reactions: SJS/TEN, MPE, DRESS, and other presentations. SJS was further divided into 3 groups according to area of skin detachment (≤1%, 2–5%, 6–9% of BSA). In the column of odds ratio, the values indicate the odds ratios, and the horizontal lines indicate 95% confidence intervals. BSA, body surface area; CI, confidence interval; DRESS, drug rash with eosinophilia and systemic symptoms; HLA, human leukocyte antigen; MPE, maculopapular exanthema; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
Among 112 patients with CBZ-induced SJS/TEN, only 5.36% (6/112) patients carried the homozygous HLA-B*15:02 allele (Fig. 4). Two patients were CBZ-induced SJS with 6–9% BSA detachment, and 4 were CBZ-induced SJS with 2–5% BSA detachment. None of CBZ-TEN cases carried the homozygous HLA-B*15:02 allele. We did not observe gene dosage effect on the disease severity.
Fig. 4HLA-B*15:02 gene dosage effect on the severity of carbamazepine (CBZ)-induced SJS/TEN. The numbers of patients with homozygous (labeled by black color), heterozygous (gray lines) or non-HLA-B*15:02 alleles (white color) are shown in the 5 subgroups of CBZ-SJS/TEN with different disease severity, according to the BSA of epidermal detachment. 6 SJS patients carry homozygous HLA-B*15:02 allele (black color), and 13 SJS patients (7 subjects with 2–5% BSA skin detachment, and 6 subjects with < 1% BSA skin detachment) do not carry HLA-B*15:02 allele (white color). In the group of CBZ-TEN, two cases were died, and one of the cases had the secondary exposure. HLA, human leukocyte antigen; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
3.7 Drug dose effect on the severity of CBZ-induced SJS/TEN
The average daily drug dose and cumulative dose of CBZ-SJS/TEN were 299.4 ± 141.1 mg/day and 3796 ± 2748.9 mg, respectively, with a latent period of 13.1 ± 7.5 days (Fig. 5). The highest daily dose taken was found in the group of SJS with 6–9% BSA detachment (375 ± 109 mg/day), while the highest cumulative dose was seen in the SJS 2–5% detachment group (4467.5 ± 3311.8 mg) (Fig. 5). A tendency for lower daily dose and cumulative dose intake was observed in the TEN group (240 ± 166.6 mg/day, 1840 ± 852.3 mg), with a shorter duration of the latent period (8.6 ± 4.5 days) (Fig. 5). Comparing the data of CBZ-SJS with <10% BSA detachment (3984.6 ± 843.2 mg), the cumulative dose of CBZ-TEN group (1840 ± 852.3 mg) was statistically significantly lower (P = 0.0013, by Student's t-test). In our serial patients, the overall mortality rate of CBZ-induced SJS/TEN was 1.79% (2/112). Two died cases were TEN with maximum BSA detachment (∼90%). One case (27 y/o, male) died after 3 days of the secondary exposure of CBZ (daily dosage: 400 mg), though he developed SJS only at the first exposure of CBZ. The other (40y/o, female) died after 13 days of her first usage of CBZ (daily dosage: 200 mg).
Fig. 5Drug dose-effect and latent periods on the severity of carbamazepine (CBZ)-SJS/TEN. (a) Average daily dosage of the 5 subgroups of CBZ-SJS/TEN; (b) average accumulation dosage of the 5 subgroups of CBZ-SJS/TEN; and (c) latent period of the 5 subgroups of CBZ-SJS/TEN. All values are expressed as mean ± SD, upper and lower limit. The thick horizontal lines indicate the mean dosage/days of each subgroup. The vertical bar indicated the standard deviation (SD). The thin horizontal lines indicate the upper and lower limit data of the each group. BSA, body surface area; CBZ, carbamazepine; SJS, Stevens–Johnson syndrome; TEN, toxic epidermal necrolysis.
3.8 Relationship between HLA-B*15:02 allele and the clinical course, clinical presentations and the disease severity of CBZ-SJS/TEN
We evaluated the relationship between the HLA-B*15:02 allele and the clinical course, clinical presentation and disease severity of CBZ-induced SJS/TEN (eTable 3). Oral involvement was observed in 100% of the patients, ocular involvement in 67% (75/112) and genital involvement in 53.7% (60/112) (eTable 3). Subgroup analysis showed relatively high percentages of ocular or genital mucosa involvement in the CBZ-TEN group (eTable 3). The percentages of ocular and genital mucosa involvement correlated positively with the extent of epidermal necrolysis. However, the extent of mucosal involvement had no significant association with the HLA-B*15:02 allele (eTable 3).
4. Discussion
By this large sample size of patients with well-defined clinical phenotypes, herein, we report the strength, sensitivity, and HLA genotype–phenotype correlations in the clinical spectrum of CBZ-induced hypersensitivity reactions. HLA-B*15:02 showed strongest association with CBZ-induced SJS/TEN, correlating with the disease severity defined by the extent of the skin detachment. The sensitivity of HLA-B*15:02 for CBZ-SJS/TEN was 100% in the patients with >5% BSA involvement of skin detachment, yet decreased to 85.1% in patients with <5% BSA involvement (Fig. 1). By comparison, CBZ-induced MPE/DRESS had no association with HLA-B*15:02 allele, but linked to HLA-A*31:01 (Pc = 2.7 × 10−3; OR (95% CI) = 6.86(2.4–19.9), and HLA-B*51:01 (Pc = 0.01; OR(95% CI) = 4.56(2.0–10.5)). HLA-B*40:01 showed negative association with CBZ-SJS/TEN ((Pc = 8.3 × 10−5; OR (95% CI) = 0.22(0.1–0.4)) (Table 2, Table 4, Fig. 3). These data revealed that CBZ-induced hypersensitivity reactions have HLA genotype–phenotype correlations, in which HLA-B*15:02 allele has the highest probabilities for testing CBZ-SJS/TEN, and HLA-A*31:01 for CBZ-MPE/DRESS in Han Chinese. No gene dosage or CBZ drug dosage effect on the severity of CBZ-induced hypersensitivity reactions was observed.
In 2006, we first identified that HLA-A*31:01 was associated with CBZ-MPE, and CBZ-MPE/DRESS in Han Chinese [
]. Recently, HLA-A*31:01 was shown to be associated with all types of CBZ-induced hypersensitivity reactions in the populations of Europeans, Japanese and Korean by genome-wide association approach [
Genome-wide association study identifies HLA-A*3101 allele as a genetic risk factor for carbamazepine-induced cutaneous adverse drug reactions in Japanese population.
]. In this study, we analyzed the 194 cases and found that HLA-A*31:01 links to CBZ-induced non-blistering hypersensitivity reactions: MPE/DRESS (P = 1.6 × 10−4; Pc = 2.7 × 10−3; OR (95% CI) = 6.86(2.4–19.9), but not CBZ-induced blistering reactions: SJS/TEN, in Han Chinese. In this study, we also found HLA-B*51:01 had significant association with CBZ-induced MPE/DRESS. We examined the possibility of haplotype between HLA-A*31:01 and HLA-B*51:01, and found that the haplotype frequencies of HLA-A*31:01 and HLA-B*51:01 are very low in Chinese as well as in different populations [
]. The association of CBZ-MPE/DRESS with either HLA-A*31:01 or HLA-B*51:01 may imply different pathogenesis role of HLA in the hypersensitivity reactions. Regarding the negative association of HLA-B*40:01 and CBZ-SJS/TEN, it could be due to the dominant effects (higher frequency) of HLA-B*15:02 in the CBZ-SJS/TEN patients and higher allele frequency of HLA-B*40:01 in the Taiwanese general population [
]. Whether HLA-B*40:01allele provided the protective effects on CBZ-SJS/TEN needed to be further evaluated.
As the data shown here that HLA-B*15:02 is not associated with CBZ-induced MPE/DRESS, the performance of HLA-B*15:02 genetic test only shall fail to prevent all types of CBZ-induced hypersensitivity reactions. Clinicians should carefully interpret the results of genetic test, and a close monitor of early clinical presentations of CBZ-hypersensitivity is still required, since DRESS and even severe MPE may also cause patients’ suffering and need hospitalization. In addition, SJS with very limited BSA of skin detachment may also cause severe mucosal involvements (especially for eye mucosa), and more severe late ocular sequelae, such as chronic keratitis, ectropion and even corneal scarring leading to blindness [
]. In addition, as HLA-B*15:02 is not 100% associated with CBZ-SJS/TEN, a prompt stop of CBZ administration when patients developing early danger signs of hypersensitivity reactions is necessary, even for patients with negative results of HLA-B*15:02 genetic test in Han Chinese. A combined test for the three risk alleles: HLA-B*15:02/HLA-B*51:01/HLA-A*31:01 may be considered in Han Chinese, and the cost-effectiveness of the combined genetic test need further evaluation.
In this study, more than 90% of patients developed CBZ-SJS/TEN within 1 month, and the average latent period of SJS/TEN caused by CBZ was 13.1 days, suggesting that close monitor of potential mucocutaneous reactions is needed for patients taking CBZ during the first one to two months. Re-challenging an individual with the culprit medication has been known to cause a shorter latent period with severe manifestations of SJS/TEN due to the immune memory response [
]. Interestingly, we also found that the latent period and the average cumulative dose were shorter and lower in the group of CBZ-TEN when comparing to CBZ-SJS. Among our CBZ-TEN patients, one patient had the secondary exposure of CBZ and rapidly developed TEN within those 3 days with BSA nearly 100% skin detachment and multiple organ failure. In addition, we found that patients with severe BSA involvement had a higher frequency of eye mucosa involvement, which is consistent with the observation of the previous report [
]. Our findings suggest that the previous sensitized reaction, or the variable immune memory response of different individuals, rather than the genetic or drug dosage effects, may account for the spectrum of clinical severity of CBZ-induced hypersensitivity reactions.
In addition to HLA-B*15:02, other members of HLA-B75 family, such as HLA-B*15:08, HLA-B*15:11, HLA-B*15:21, and HLA-B*15:31, have also been reported to associate with CBZ-SJS/TEN in different populations [
]. In our study, 2 cases of CBZ-SJS carried HLA-B*15:11 instead of B*15:02, and HLA-B*15:11 showed a weak association with CBZ-induced SJS/TEN (P = 0.007; Pc = 0.21; OR(95% CI) = 61.52(2.8–1359.3) (Table S2). These data imply that risky alleles belonging to HLA-B75 serotype may encode proteins sharing similar conformation with HLA-B*15:02, which can react to CBZ to trigger the immune response. Our functional study supports this hypothesis, as HLA-B*15:02 recombinant protein and members of the HLA-B75 family possess higher binding affinity toward CBZ and its structural analogs to trigger the cytotoxic T lymphocytes-mediated immune reactions [
In conclusion, this study revealed the clinical spectrum and HLA genotype–phenotype correlations in CBZ-induced hypersensitivity reactions. In particular, the HLA-B*15:02 allele showed increased sensitivity, OR and probability for CBZ-SJS/TEN with the more BSA involvement of skin detachment. HLA-B*40:01 has a negative association with CBZ-SJS/TEN. By comparison, CBZ-MPE/DRESS showed no association with HLA-B*15:02, but links to HLA-A*31:01 and HLA-B*51:01 in Han Chinese. No HLA gene dosage or CBZ dosage effect on the severity of CBZ-induced hypersensitivity reactions was observed. This study suggests that even with the advantage of a HLA genetic test for preventing adverse reactions, the HLA genotype–phenotype relationship and the results of the test need to be carefully interpreted for CBZ-induced hypersensitivity reactions.
Acknowledgements
This work was supported by grants from the National Science Council, Taiwan (99-3114-B-182A-001; 101-2320-B-010-072-MY3; NSC101-2321-B-010-027) and grants from Chang-Gung Memorial Hospital (BMRPG-290011, CMRPG-290051∼3). We thank Y.Y.W. for the maintenance of the database and Prof. T.t.K. for reading the pathology slides.
Appendix A. Supplementary data
The following are the supplementary data to this article:
Association of HLA-B*1502 allele with carbamazepine-induced toxic epidermal necrolysis and Stevens–Johnson syndrome in the multi-ethnic Malaysian population.
Genome-wide association study identifies HLA-A*3101 allele as a genetic risk factor for carbamazepine-induced cutaneous adverse drug reactions in Japanese population.
ALDEN, an algorithm for assessment of drug causality in Stevens–Johnson Syndrome and toxic epidermal necrolysis: comparison with case-control analysis.
Correlations between clinical patterns and causes of erythema multiforme majus, Stevens–Johnson syndrome, and toxic epidermal necrolysis: results of an international prospective study.
00337-X&id=gr1.jpg){kind=link}
00337-X&id=gr2.jpg){kind=link}
00337-X&id=gr3.jpg){kind=link}
00337-X&id=gr4.jpg){kind=link}
00337-X&id=gr5.jpg){kind=link}
00337-X&id=mmc3.jpg){kind=link}