Identification of risk factors for carbamazepine-induced serious mucocutaneous adverse reactions: A case-control study using data from spontaneous adverse drug reaction reports.

OBJECTIVES: To identify risk factors other than genetic for severe carbamazepine-induced mucocutaneous reactions, that is, SJS, TEN, and exfoliative dermatitis (ED).
MATERIALS AND METHODS: We did a case-control study using data from the Swedish national database of spontaneously reported adverse drug reactions (ADRs). We selected all patients who had been reported from January 1, 1965 to March 31, 2010 as having experienced SJS (n = 78), TEN (n = 6), or ED (n = 8), and assessed as at least possibly related to carbamazepine. We also included diagnoses possibly representative of early signs of these serious conditions, that is, erythema multiforme (EM, n = 34) and scaly rash (n = 13). We compared data on demographics, drug treatment, and clinical features for these patients (cases, n = 139) with those from patients who had experienced any other type of ADR from carbamazepine during the same time period (controls, n = 887).
RESULTS: After adjustment for multiple comparisons, alcohol abuse was statistically significantly more common among cases than controls (34.5% vs 8.7%, odds ratio 5.5 [95% confidence interval 3.6-8.4], P = 3.14 × 10(-14) ). The same was seen for SJS and EM individually.
CONCLUSION: Alcohol abuse is a possible risk factor for serious carbamazepine-induced mucocutaneous reactions.

INTRODUCTION

Adverse drug reactions (ADRs) are among the leading causes of morbidity and mortality. [1] Most are dose-dependent and pharmacologically predictable (type A reactions). [2] Some, however, have no known pharmacological cause (type B reactions). This latter type of ADRs are nonpredictable, commonly serious and lead more frequently than type A reactions to withdrawal of drugs from the market. [3] Notable examples include the severe mucocutaneous eruptions Stevens–Johnson syndrome (SJS), toxic epidermal necrolysis (TEN), exfoliative dermatitis (ED), and erythema multiforme (EM) majus.

SJS and TEN typically develop due to specific medications, although other triggers such as infections with Mycoplasma pneumoniae and Herpes simplex virus are also possible. [4] SJS and TEN commonly present with fever and malaise, developing into cutaneous erythema with blister formation of various extent and hemorrhagic erosions of mucous membranes, for example, stomatitis, balanitis, colpitis, severe conjunctivitis, and blepharitis. [5] There is still some debate whether SJS and TEN represent different diseases, but most experts consider them a single disease of different severity, [15] with TEN being the more severe with epidermal detachment of >30% of the body surface area and SJS the less severe defined as <10% detachment. [6] SJS and TEN are extremely rare, with incidence rates of 1-6 and 0.4-1.2 cases/million persons, respectively, and are associated with mortality rates of up to 5% and 40%, respectively. [1]

SJS and TEN have long been considered to be variants of EM majus, but are nowadays considered as separate conditions due to differences in clinical presentation and potential causes. [57] Virtually no patient with EM majus dies as a consequence of the condition. Although rarely drug-induced, ED, characterized by erythema and moderate to extensive scaling with involvement of total, or near total, body surface, share many of the clinical features of SJS and TEN. [8] Whether or not ED is part of the same spectrum as SJS and TEN is unclear. It has been estimated to occur among 1-2/100,000 persons. [8]

Pharmacovigilance studies have shown that SJS, TEN, and ED can be induced by up to 200 different drugs, although certain drugs including carbamazepine, phenytoin, and allopurinol are the most common ones. [18] Whereas very little is known about the pathogenesis of ED, the mechanism by which SJS and TEN are induced is likely to be immune-mediated, [49] although not fully understood. Candidate gene studies have shown that susceptibility to SJS/TEN has a significant genetic basis, while ED has not been studied in this respect. In genetic association studies of alleles within the HLA class I region, HLA-B*1502 has been shown to be strongly associated with carbamazepine-induced SJS/TEN in Han Chinese individuals. [10] In Caucasians, HLA-B*1502 is rare and a similar association in this population has not been found. Instead, HLA-A*3101 has been identified as a predictor of carbamazepine-induced cutaneous ADRs such as SJS and TEN. [11] However, the presence of the HLA-A*3101 allele is neither necessary nor sufficient for their development.

The rarity of the HLA-B*1502 allele in Europeans, coupled with the fact that the association with HLA-A*3101 is not sufficient to explain the pathogenesis of carbamazepine-induced serious cutaneous ADRs highlights the need for further studies. We therefore aimed to identify possible risk factors other than genetic for severe mucocutaneous eruptions, that is, SJS, TEN, and ED, associated with carbamazepine in a predominantly Caucasian population. We also present a new approach for using databases of spontaneous ADR reports for the detection of possible risk factors for ADRs.

MATERIALS AND METHODS

We did a case-control study using data from the Swedish database of spontaneously reported ADRs (Swedish Drug Information System, SWEDIS). [12] This database contains reports sent by Swedish physicians to the Swedish Medical Products Agency (MPA) since 1965. As of March 31, 2010, there were in total 103,235 reports in the database. In SWEDIS, a Swedish dictionary is used for coding ADRs, built on a three-level hierarchical structure developed by the MPA (Swedish adverse reaction terminology, SWE-ART). The first level is the system organ class, followed by group terms and finally preferred terms. The dictionary holds a little over 1000 preferred terms. The information in a report consists of patient demographics, reported ADRs, medication, and a case narrative, often accompanied by copies of medical records. All reports are reviewed by clinical assessors at the MPA and a causality assessment is made. All drugs, both suspected and concomitant, are coded using the World Health Organization (WHO) Collaborating Centre for Drug Statistics Methodology International Anatomical Therapeutic Chemical (ATC) classification. [13]

For the present study, we selected all patients who had been reported from January 1, 1965 to March 31, 2010 as having experienced reactions coded in SWE-ART as SJS (n = 78), TEN (n = 6), or ED (n = 8), assessed as at least possibly related to carbamazepine treatment (cases, n = 92). We also included SWE-ART terms of reactions, which can be difficult to differentiate from early signs of these serious conditions, that is, EM (n = 34) and scaly rash (n = 13). We compared data on demographics, drug treatment and clinical features for these patients (cases) with those from patients who had experienced any other type of ADR during the same time period (controls). From the initial complete sample of 910 controls, we excluded reports of congenital malformations and miscarriage (n = 23) as these are reflective of adverse effects on the fetus and not primarily the patient, arriving at a number of 887 controls. Otherwise there were no restrictions for inclusion.

For each patient, we entered data (age, sex, dose of carbamazepine, time to onset of the ADR, concomitant drugs and their ATC-codes, indication for treatment with carbamazepine and comorbidities) into an Excel spreadsheet (Microsoft Corporation, USA, v14.0). Diseases, that is, indication for treatment and comorbidities, were coded according to their closest matching ICD-code (ICD 10). [14] We also grouped diseases of similar type [Table 1] and drugs belonging to the same class [Tables 2 and 3]. Drugs/drug-classes and diseases, which were assessed as unlikely to be associated with SJS/TEN/ED, as well as diseases that were too poorly defined to be categorized accurately, were excluded [Table 4]. We further excluded drugs/drug-classes and diseases occurring among less than six patients in the complete data-set [Table 3], thereby arriving at a final number of 144 variables for analysis [Table 2].

Table 1

Grouping of diseases of similar type

Table 2

Variables included in the analyses. ATC-codes N02AA59 (codeine/acetylsalicylic acid/caffeine), M03BB53 (acetylsalicylic acid/dextropropoxyphene/chlorzoxazone) and M01AE51 (ibuprofen/codeine) were placed in both of the drug groups M01A+N02BA+B01AC06+N02AA59+M03BB53+M01AE51 and N02A+M03BB53+M01AE51. ATC-code B01AC06 (acetylsalicylic acid) was also placed in two groups, B01AC and M01A+N02BA+B01AC06+N02AA59+M03BB53

Table 3

Variables excluded from analyses (occurring among less than six patients)

Table 4

Excluded drugs/drug groups unlikely to be associated with SJS/TEN/ED, and excluded conditions too poorly defined

Differences between cases and controls regarding age, sex, dose of carbamazepine, time to onset of the ADR, and for each variable in Table 2 were assessed for SJS, TEN, ED, EM, and scaly rash both separately and as a group. Fisher's exact test was used when comparing proportions and Student's t-test for mean values. We used Bonferroni's correction to adjust for the 864 comparisons that were made (144 × 6), which gave a two-tailed P value of 5.787 × 10-5 for statistical significance with 80% power. When comparing the mean daily dose of carbamazepine between groups, we excluded extreme outliers, that is, cases of intentional self-intoxication with high doses (n = 3, doses ranging from 6000 to 20000 mg), young children below the age of 12 years (n = 80), and cases where information about the dose was likely to be erroneous (n = 3, daily doses ranging from 0.2 to 40 mg). For time to onset of the ADR, we similarly excluded extreme outliers (n = 6, time ranging from 7300 to 22,920 days). As the distribution of time to onset of the ADR was markedly skewed we ln-transformed this variable. All calculations were performed in JMP 7.0.1 (SAS Institute Inc., Cary, North Carolina, USA).

To test the validity of the data and the appropriateness of the study methodology, we also investigated whether known risk factors for carbamazepine-induced hyponatremia and SIADH, that is, high age, female gender, high dose, and concomitant treatment with antidepressants or diuretics (loop-diuretics and thiazides), [1516171819] would be more common among cases than controls, as expected. In the present dataset, there were 38 cases of hyponatremia and 14 of SIADH (terms according to SWE-ART).

Ethics approval

The study was approved by the local ethics committee (Sweden, Uppsala, 2012/200).

RESULTS

Patient characteristics in the complete data-set are shown in Table 5. The mean age was 48.9 years and there were slightly more women than men (51.3% vs 48.7%). The most common indication for treatment with carbamazepine was epilepsy, followed by trigeminal neuralgia and alcohol-related mental disorders. About 70% of all reports had been classified as serious. The most commonly involved system organ class was skin and subcutaneous tissue disorders (37.2%).

Table 5

Characteristics of the 1026 patients in the complete data-set

Differences between cases and controls for each studied variable are shown in Tables 6–11. When all investigated skin reactions (SJS, TEN, ED, EM, and scaly rash) were grouped [Table 6], alcohol-related mental and behavioral disorders were statistically significantly more common among cases than controls (34.5%, n = 48 vs 8.7%, n = 77; odds ratio (OR)=5.5 [95% confidence interval (CI) 3.6-8.4], P = 3.14 × 10-14 ), as was the disease group substance/alcohol abuse (37.4%, n = 52 vs 9.8%, n = 87; OR = 5.5 [3.7-8.3], P = 5.99 × 10-15 ). Mean time to onset was shorter among cases than controls (67 vs 127 days). The same pattern was evident for SJS [Table 7] and EM [Table 10] individually, although time to onset was not significantly different from the control group in the latter case. No statistically significant differences were seen for TEN [Table 8], ED [Table 9] and scaly rash [Table 11] individually, although for the latter, ORs similar to those for SJS and EM were seen regarding alcohol-related mental and behavioral disorders. For all of the 48 cases with any of the investigated skin reactions and who were classified with the ICD-10 code alcohol-related mental and behavioral disorders, as well as for the corresponding 77 controls, this ICD-10 classification had been made due to alcoholism. According to the information contained within the 48 reports, the indication for treatment with carbamazepine for 40 was alcoholism. The precise condition was not always stated but alcohol withdrawal was commonly described. In Sweden, alcohol withdrawal is the only approved indication related to alcoholism for carbamazepine. [20] The indications for treatment among the remaining eight were Wernicke-Korsakoff syndrome in one case, epilepsy in four, and unspecified convulsions in three. Concomitant substance abuse was mentioned in the case narratives for 4 of the 48 cases (1 each of amphetamine and morphine, and 2 of unspecified drugs).

Table 6

Differences in patient demographics, daily dose of carbamazepine, time to onset of adverse reaction, medical history and concomitant drug treatment between cases and controls-all investigated skin reactions (Stevens-Johnson syndrome, toxic epidermal necrolysis, exfoliative dermatitis, erythema multiforme and scaly rash). Drugs/drug groups are defined by their ATC-codes. Individual diseases are classified according to ICD-10

Table 11

Differences in patient demographics, daily dose of carbamazepine, time to onset of the adverse reaction, medical history and concomitant drug treatment between cases and controls-scaly rash. Stevens-Johnson syndrome, toxic epidermal necrolysis, exfoliative dermatitis and erythema multiforme are excluded. Drugs/drug groups are defined by their ATC-codes. Individual diseases are classified according to ICD-10

Table 7

Differences in patient demographics, daily dose of carbamazepine, time to onset of the adverse reaction, medical history and concomitant drug treatment between cases and controls- Stevens-Johnson syndrome. Toxic epidermal necrolysis, exfoliative dermatitis, erythema multiforme and scaly rash are excluded. Drugs/drug groups are defined by their ATC-codes. Individual diseases are classified according to ICD-10

Table 10

Differences in patient demographics, daily dose of carbamazepine, time to onset of the adverse reaction, medical history and concomitant drug treatment between cases and controls-erythema multiforme. Stevens-Johnson syndrome, toxic epidermal necrolysis, exfoliative dermatitis and exfoliative rash are excluded. Drugs/drug groups are defined by their ATC-codes. Individual diseases are classified according to ICD-10

Table 8

Differences in patient demographics, daily dose of carbamazepine, time to onset of the adverse reaction, medical history and concomitant drug treatment between cases and controls-toxic epidermal necrolysis. Stevens-Johnson syndrome, exfoliative dermatitis, erythema multiforme, and exfoliative rash are excluded. Drugs/drug groups are defined by their ATC-codes. Individual diseases are classified according to ICD-10

Table 9

Differences in patient demographics, daily dose of carbamazepine, time to onset of the adverse reaction, medical history and concomitant drug treatment between cases and controls-exfoliative dermatitis. Stevens-Johnson syndrome, toxic epidermal necrolysis, erythema multiforme and exfoliative rash are excluded. Drugs/drug groups are defined by their ATC-codes. Individual diseases are classified according to ICD-10

Table 12 shows differences in age, gender, daily dose of carbamazepine, and concomitant treatment with antidepressants or diuretics between patients with hyponatremia/SIADH and controls. Patients who had experienced hyponatremia or SIADH had a higher mean age than controls (64.0 vs 48.1 years, P = 7.87 × 10-7 ) were more often women (78.4% vs 49.9%, P = 7.24 × 10-5 ), and were more often on concomitant treatment with antidepressants (26.9% vs 7.7%, P = 5.73 × 10-5 ) or diuretics (23.1% vs 6.2%, P = 1.23 × 10-4 ). However, there was no difference in daily dose of carbamazepine.

Table 12

Differences in known risk factors for hyponatremia/SIADH (SWE-ART terms) during treatment with carbamazepine. Drugs/drug groups are classified by their ATC-codes

DISCUSSION

In this case-control study of spontaneous ADR reports for carbamazepine sent from healthcare professionals to the Swedish drug regulatory authority, we identified alcohol abuse as well as substance/alcohol abuse as possible risk factors for serious mucocutaneous reactions. We have also shown that the study methodology can be used to identify possible risk factors for ADR in databases of spontaneous ADR reports. The identified associations were seen for SJS and EM but not for TEN, ED, and scaly rash. Whether this reflects a true difference between the reactions or simply a difference in sample size – SJS and EM made up 80% of all investigated skin reactions – is not possible to determine. The possibility of an association between alcohol and/or substance abuse and any type of skin reaction was assessed as unlikely as there were no statistically significant differences for these variables between patients with dermatologic ADRs other than SJS, TEN, ED, EM and scaly rash, and patients with nondermatologic ADRs [Table 13].

Table 13

Differences in variables reflective of alcohol/substance abuse between patients with skin reactions (SWE-ART system organ class skin and subcutaneous tissue disorders) other than Stevens-Johnson syndrome, toxic epidermal necrolysis, exfoliative dermatitis, erythema multiforme and scaly rash, and patients with nondermatologic ADRs. Drugs/drug groups are defined by their ATC-codes. Individual diseases are classified according to ICD-10

Our study has limitations. One concerns the information contained within a report. Bias may have been introduced if the extent and detail of clinical data differed between cases and controls, that is, the reporter or the assessor handling the ADR report may have decided to include or retrieve more data for serious mucocutaneous reactions than less severe or nonserious and common ADRs. Such information may have included alcohol use. Indeed, in one of the case narratives for a report of SJS during 1986, the assessor at the MPA had made a comment that “these cases often involve a history of alcohol problems, then withdrawal, and then treatment with carbamazepine”. However, none of the other case narratives mentioned a suspected link with alcohol abuse. A second limitation is the known association between alcohol and other substance abuse. Because of this strong association, we grouped the individual ICD-10 codes mental and behavioral disorders due to use of alcohol with mental and behavioral disorders due to multiple drug use and use of other psychoactive substances, into the disease group substance/alcohol abuse. The association with serious mucocutaneous reactions was slightly stronger for the disease group than for the individual diseases, making it somewhat problematic to pinpoint the associated component. However, as there was no statistically significant association with mental and behavioral disorders due to multiple drug use and use of other psychoactive substances, and since this was co-present with alcohol abuse in only very few of the cases with serious mucocuaneous reactions, we believe that the component most likely to be associated with the studied ADRs is alcohol abuse. A third limitation is the lack of information for most cases whether alcohol abuse was ongoing, had occurred in the past, or if a patient was experiencing withdrawal symptoms. The last limitation is the inability of epidemiological studies to establish cause-effect relationships. Thus, alcohol abuse may simply be a marker of another or other, unknown risk factor (s). Such factors might include malnutrition or immunological effects due to overconsumption of alcohol.

To our knowledge, this is the first study to suggest alcohol abuse as a risk factor for carbamazepine-induced serious mucocutaneous reactions. Illing et al. recently showed that abacavir, a drug strongly associated with acute hypersensitivity syndrome among carriers of the HLA-B*5701 allele, binds to HLA-B*5701 and changes the shape and chemistry of the antigen-binding cleft, thereby altering the repertoire of endogenous peptides that can bind HLA-B*5701. [21] In this way, abacavir guides the selection of new endogenous peptides, inducing activation of abacavir-specific T-cells with driving of polyclonal CD8 T-cell activation and a systemic reaction manifesting as acute hypersensitivity syndrome. It was also shown that carbamazepine binds to HLA-B*1502, producing alterations in the repertoire of presented self peptides, and that this is the likely mechanism underlying reactions such as SJS. The same mechanism is therefore likely to be involved in the association between HLA-A*3101 and the risk of carbamazepine-induced serious cutaneous reactions in Europeans. However, as the presence of the HLA-A*3101 allele is not sufficient to explain the pathogenesis of these reactions, other mechanisms must also be involved. Such a factor might be alcohol. We have found two published case reports suggesting an association between alcohol abuse and serious cutaneous reactions to carbamazepine when given for treatment of withdrawal symptoms, [2223] although heroin was co-suspected in one case. Another case report also describes the onset of SJS in a patient with alcohol abuse following treatment of withdrawal symptoms with carbamazepine, although a possible link with alcohol abuse was not suspected. [24] A further three reports of SJS in patients with alcohol abuse have been published for other drugs, that is, lamotrigine, [25] chlarithromycin and/or disulfiram, [26] and phenytoin. [27] One case report also describes recurrent EM in a patient on therapy with ciprofloxacin after ingestion of alcohol. [28]

The underlying mechanism whereby alcohol abuse might increase the risk of carbamazepine-induced serious cutaneous reactions will remain speculative. Like the majority of secreted, cell surface, and immune-related molecules, HLA proteins contain posttranslational glycan modifications that are initiated in the endoplasmic reticulum. [29] Such glycans can stabilize the conformation of the protein structure both locally and at distant protein domains. In disease settings, however, these glycan structures can be altered significantly through factors such as altered expression patterns of glycosidases and glycosyltransferases in the golgi apparatus, altered availability of carbohydrate building blocks, and increased expression of cellular proteins that may overwhelm the limits of the secretory pathway. Inflammatory stimuli, for instance, can alter the glycan branching pattern and levels of sialylation of cell bound and secreted glycoproteins. Perhaps chronic alcohol ingestion modifies HLA protein dynamics as well as the structure–function relationship through such a mechanism.

CONCLUSION

In conclusion, we have identified alcohol abuse as a possible risk factor for carbamazepine-induced serious mucocutaneous reactions. This association needs confirmation in other data-sets. We have also shown that databases of spontaneous ADR reports may still be under-utilized, and that case-control studies as described within this report could be used for the detection of possible risk factors of ADRs.