I. Epidemiology of Cutaneous Drug Reactions

Top Previous Next References

Although the true incidence of adverse drug reactions (ADRs²) is difficult to quantify, there is abundant evidence that cutaneous drug reactions (CDRs) are among the most frequent adverse events in patients receiving drug therapy. In a study examining the incidence of CDRs in spontaneous ADRs reported in Italy, over 30% of all reported ADRs were cutaneous in nature (Naldi et al., 1999). This is higher than other studies, where CDRs comprise 10 to 20% of reported ADRsstill a substantial fraction of all reactions (Stewart et al., 1979; Faich et al., 1987; van der Linden et al., 1998). Differences in methodology for data collection (e.g., spontaneous reporting versus chart review) may explain the variability in reported frequencies.

Among all hospitalized patients, the incidence of CDRs has been found to range from 1 to 3% (Arndt and Jick, 1976; Stewart et al., 1979; Bigby et al., 1986). Numerous risk factors, including infection with the human immunodeficiency virus (Battegay et al., 1989; Coopman and Stern, 1991; Bayard et al., 1992; Spira et al., 1998; Gonzalez-Martin et al., 1999), infectious mononucleosis (Patel, 1967; Pullen et al., 1967; Nazareth et al., 1972; van der Linden et al., 1998), female sex (Bigby et al., 1986; Naldi et al., 1999; Fattinger et al., 2000), and age (Naldi et al., 1999; Ibia et al., 2000), have been identified. The highest reported frequency of CDRs has consistently been found to be with antimicrobial agents, for which there is also limited data regarding the incidence of CDRs in ambulatory patients. In a population of 13,679 Dutch patients who received a prescription for an antimicrobial agent, the frequency of CDRs was approximately 1% (van der Linden et al., 1998). The most frequent reactions were observed in patients receiving a trimethoprim-sulfonamide combination (2.1%), fluoroquinolones (1.6%), and penicillins (1.1%). A higher frequency of CDRs (7.3%) was recently noted in an ambulatory pediatric population receiving penicillins, sulfonamides, or cephalosporins (Ibia et al., 2000). The percentage of children receiving cefaclor, sulfonamides, penicillins, or other cephalosporins who exhibited a rash during treatment was 12.3, 8.5, 7.4, and 2.6%, respectively. The odds ratio of development of a rash was higher in children less than 3 years of age than other groups.

Determination of the true incidence of CDRs is difficult due to imprecise diagnostic criteria. Causality assessment varies among reported studies and is limited by the ethical constraints of rechallenging patients with a drug that may evoke a life-threatening or seriously disabling reaction. As skin rashes can often occur in the presence of bacterial or viral infection in the absence of drug therapy, determination of causality in suspected CDRs with antimicrobial agents is especially problematic. Several studies have suggested that Bayesian approaches may serve as a useful aid in the differential diagnosis of a suspected CDR, but there is no "gold standard" against which to compare such methods for accuracy (Kwok et al., 1994; Lanctot et al., 1994).

Based on the frequency of drug ingestion and the incidence of CDRs for inpatients from the Boston Collaborative Drug Surveillance Program, Levenson and colleagues (1991) estimated in 1991 that 2.25 million patients in the United States experience CDRs each year. As the frequency of drug ingestion has significantly increased since that date, the anticipated number of patients experiencing such reactions is considerably higher than this previous estimate. Because it has recently been estimated that between 5 and 9% of all hospital costs are related to ADRs (Moore et al., 1998), CDRs clearly represent a significant burden on the health care system, in addition to being a frequent reason for cessation of otherwise effective drug therapy in patients.

	II. Clinical Morphology of Cutaneous Drug Reactions

Top Previous Next References

Morbilliform rashes are common CDRs characterized by primary skin lesions of fine pink macules and papules that may become confluent. Typically, lesions begin on the trunk and pressure-bearing areas and progress symmetrically to cover large areas of the body (Fig. 1). Lesions usually begin within 1 to 2 weeks of starting a medication and fade 1 to 2 weeks following cessation.

View larger version (130K): [in this window] [in a new window]   Fig. 1.   Morbilliform eruption. The patient developed a pruritic maculo-papular eruption on the torso after receiving trimethoprim-sulfamethoxazole.

Histologic examination reveals a lymphocytic interface dermatitis with vacuolar changes at the dermal-epidermal junction and papillary dermal edema and eosinophils. Dyskeratotic cells may be found along the dermal-epidermal junction (Crowson and Magro, 1999).

Because CDRs, particularly morbilliform reactions, are so common and because they may be seen with many different classes of medications, identification of the causative agent can be a significant challenge (Table 1) (Breathnach, 1998; Crowson and Magro, 1999). Roujeau and Stern (1994) describe several criteria that may be helpful in defining a CDR: 1) other causes for the eruption, such as viral exanthem, should be excluded; 2) a temporal relationship between drug use and onset of the rash should exist; 3) improvement should be noted following drug cessation; 4) reactivation upon rechallenge of the drug should be noted; and 5) the cutaneous reaction is known to be associated with the drug in question. Drug Eruption Reference Manual (Litt, 1999) is an excellent source of detailed listings of the reported type and frequency of eruptions associated with a particular medication.

A careful history and research into the causative agent will allow prompt withdrawal of the offending agent and can prevent a patient from being falsely labeled with multiple drug allergies. There is some evidence that certain individuals are more likely to develop reactions to multiple medications, but a "multiple drug allergy syndrome" has not been clearly established (Gruchalla, 2000). The labeling of patients with multiple drug allergies to antibiotics may be contributing to the overuse of broad-spectrum antibiotics and increasing drug resistance.

Urticaria is the second most common CDR (Table 1) (Breathnach, 1998). The hallmark features of an urticarial eruption are the development of pink wheals on the skin with accompanying pruritus. Angioedema is the corresponding tissue reaction of urticaria occurring deeper in the subcutaneous tissue. The clinical appearance of drug-induced urticaria is indistinguishable from urticaria attributable to other causes. The pruritic wheals of urticaria arise as edematous plaques that may range from red to flesh-colored to yellow (Fig. 2). Follicular prominence may be increased as the hair follicles are anchored deeply in the dermis below the tissue edema. As lesions progress, a raised annular or serpiginous border with central clearing may be seen. As a rule, single lesions last less than 24 h, although new lesions may continue to arise. In contrast, angioedema is characterized by deep, skin-colored swelling, most commonly of the lips or eyes, that may last for several days. It may occur with urticarial lesions or arise independently.

View larger version (101K): [in this window] [in a new window]   Fig. 2.   Acute generalized urticaria secondary to ampicillin. The patient was receiving the antibiotic to treat a urinary tract infection. Depicted area is the anterior thigh.

The histology of an urticarial drug reaction is indistinguishable from that of urticaria from other causes. Interstitial dermal edema with endothelial swelling is present with variable infiltrate of lymphocytes, neutrophils, and eosinophils in the dermis.

Urticarial vasculitis is a distinct small-vessel vasculitic disease that should be differentiated from urticaria. It is characterized by painful rather than pruritic urticarial skin lesions that persist for greater than 24 h and usually resolve with postinflammatory pigmentation. Systemic findings are also common and include arthritis and abdominal or chest pain. Most cases of urticarial vasculitis are idiopathic but it may be associated with collagen vascular disease and viral infection (Odom et al., 2000).

Fixed drug eruption represents a unique CDR pattern characterized by skin lesion(s) that recur at the same anatomic site(s) upon repeated exposures to an offending agent. Most commonly, the skin lesion is a dusky erythematous macule and is usually found on the lips and genitalia, although any skin or mucosal surface may be involved. The skin lesions may be associated with a burning sensation and may be present in multiple numbers or progress to the development of central vesicles and bullae, particularly after the repeated use of an agent (Fig. 3). The skin findings may be associated with nonspecific constitutional symptoms, including fever, malaise, nausea, and vomiting.

View larger version (134K): [in this window] [in a new window]   Fig. 3.   Bullous fixed drug eruption due to pseudoephedrine. The patient presented with multiple oval erythematous patches following the use of allergy sinus medication containing pseudoephedrine. The patient had two previous episodes from different sympathomimetic medications. Several of the larger patches contained intact flaccid bullae. Bullae formation is uncommon in fixed drug eruption and may be confused with other severe blistering drug reactions, including SJS and TEN.

Fixed drug eruption usually occurs within hours of administration of the offending agent. Most commonly implicated are sulfa medications, barbiturates, and tetracycline, although many different medications have been implicated (Table 1) (Crowson and Magro, 1999; Stern and Wintroub, 1999). Certain medications also appear to have a predilection for certain anatomic locations (e.g., tetracyclines: genitalia).

Histology demonstrates a mixed inflammatory infiltrate of lymphocytes, neutrophils, and eosinophils at the dermal-epidermal junction. The epidermis contains necrotic keratinocytes. Melanin-containing macrophages in the dermis and chronic epidermal changes including acanthosis, hypergranulosis, and hyperkeratosis may be seen in older lesions (Crowson and Magro, 1999).

The differential diagnosis of fixed drug eruption includes erythema multiforme (EM), as well as Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN), in cases of disseminated or bullous fixed drug eruption. In contrast to fixed drug eruption, recurrent lesions in EM tend not to recur in the sites of previous involvement (Sowden and Smith, 1990). Unlike SJS and TEN, mucous membrane involvement is not a consistent finding in fixed drug eruption.

Fixed drug eruptions resolve spontaneously without scarring a few weeks after onset, usually with residual postinflammatory pigmentation. A non-pigmenting variant of fixed drug eruption was described by Shelley and Shelley (1987). In this variant, which is most common following the use of pseudoephedrine, lesions resolve spontaneously without evidence of postinflammatory pigmentation.

A careful history of previous drug exposures and skin lesions is integral in identification of the causative drug. Provocation testing with oral rechallenge of smaller doses of the suspected agent is the best method to determine with certainty the source of a fixed drug eruption. In cases of bullous or disseminated fixed drug eruption, however, oral challenge may be potentially dangerous. Patch tests, scratch tests, and intracutaneous testing may be helpful if positive, but are less reliable means of testing. When possible, patch testing should be performed at the site of previous eruption in conjunction with a vehicle to aid in drug absorption (Lisi and Stingeni, 1993).

The spectrum of severe cutaneous reactions, EM, SJS, and TEN, are the most feared of the CDRs. TEN occurs exclusively as a result of drug exposure and mortality reaches nearly 30%. An epidemiologic survey in France from 1981 to 1985 of private dermatologists, hospital dermatology, burn, intensive care, and infectious disease units estimated the prevalence of TEN to be 1.5 cases per million per year (Roujeau, 1987).

The prototypical lesion of EM is a targetoid dusky erythematous patch, found predominantly on the extremities, although as the name implies, many different morphologies may be observed. EM minor is attributable nearly exclusively to herpetic or mycoplasma infection, whereas EM major, considered by some to be synonymous with SJS, involves the mucous membranes and is associated with drug exposure in approximately 50% of cases. In SJS, bullae form on an erythematous base and confluent areas of skin detachment may be present (Fig. 4). In contrast to EM minor, initial lesions of SJS tend to occur on the face and trunk and are associated with a "burning" sensation or pain, an ominous sign of an impending severe reaction. At least two mucosal sites are involved with SJS. TEN is the most severe form in this reaction spectrum, occurring when large areas of skin sloughing affect greater than 30% of the total body surface area.

View larger version (121K): [in this window] [in a new window]   Fig. 4.   Erythema multiforme major/Stevens-Johnson syndrome. This HIV-infected patient developed desquamation of the oral mucosa following administration of trimethoprim-sulfamethoxazole.

Constitutional symptoms are more common with TEN, but both TEN and SJS patients may present with fever that precedes the mucocutaneous eruption by 1 to 3 days (Roujeau and Stern, 1994). The mucocutaneous findings of both SJS and TEN progress rapidly over hours to days. Lesions of the gastrointestinal and respiratory tract are not uncommon in these reactions, the latter of which may be a significant source of mortality. Conjunctival involvement may result in synechiae, persistent photophobia, visual impairment, or complete blindness (Roujeau and Stern, 1994).

Skin biopsy may be helpful in excluding other bullous dermatoses. Some authors believe that EM due to non-drug causes should be considered a distinct entity from more severe drug-induced SJS and TEN. EM is characterized by a dense dermal inflammatory cell infiltrate and keratinocyte necrosis In contrast, TEN shows complete epidermal necrosis and a sparse mononuclear cell infiltrate (Paquet and Piérard, 1997).

If a severe cutaneous reaction is suspected, immediate withdrawal of all potential offending agents is the most effective mode of therapy. Patients with extensive involvement should be cared for as a "burn patient" with fluid resuscitation, infection control measures, and nutritional support in a hospital burn-unit setting.

Antibiotics, particularly sulfonamides and penicillins, are traditionally implicated in many cases of severe drug reactions, but Roujeau's (1987) survey of drug eruptions in France from 1981 to 1985 found that NSAIDs had emerged as the more common cause of TEN. This was attributed in part to the availability of several new NSAIDs after 1980 (Roujeau, 1987). A list of the agents most frequently associated with SJS and TEN is given in Table 1 (Chan et al., 1990; Roujeau et al., 1990, 1995).

Rechallenge with an offending agent may result in a more severe phenotype and should not be performed following a severe drug reaction (Roujeau and Stern, 1994). Although the onset of severe cutaneous eruptions is typically one to three weeks after initiation of therapy, it may be significantly shorter upon rechallenge. Patch testing has been advocated in cases of severe CDR involving multiple candidate drugs and in which the patient may require future treatment with one of the possible agents (Gebhardt and Wollina, 1997). Difficulties arise in terms of finding an appropriate preparation of the drug suitable for patch testing and the potential risk of eliciting a severe adverse reaction.

	III. Mechanisms of Cutaneous Drug Reactions

Top Previous Next References

Being among the most frequent adverse reactions to drugs, cutaneous reactions represent common ground when it comes to patients being labeled as "allergic" to an offending (real or presumed) agent. This label, however, fails to differentiate between the many mechanisms by which drugs can elicit adverse reactions manifested in the skin; some of which may not be immune-mediated. An understanding of the various mechanisms of CDRs is essential for the identification of predisposing factors, likelihood of cross-reactivity, expected time course, and advisability of subsequent or continued therapy with the offending or related agents.

The proper terminology for reactions to drugs that are believed to be immune-mediated is an area of debate. Immunologists frequently define the term allergy as "disease mediated by the immune system to an otherwise innocuous agent". Although this definition may well describe the reaction to foodstuffs and environmental allergens, strict adherence to this definition would exclude drugs since they are anything but innocuous. Some object to the use of the term hypersensitivity for such reactions, because it "invites confusion with other kinds of adverse reactions" (Pratt, 1990). Moreover, a patient who exhibits a pharmacologic response to a drug (that is not rooted in an immune response) at a dose considerably lower than that needed to elicit the same response in the average individual, could properly be described as "hypersensitive" to that drug. Recognizing the ambiguity in the terminology frequently used for such reactions, it is probably best to refer to these reactions in general as idiosyncratic drug reactions. This classification is particularly helpful because it describes the unexpected and low incidence of these reactions while recognizing that the mechanism of many such reactions are still poorly understood. However, even this term has limited usefulness. For example, can a reaction that occurs in over 40% of a patient population (such as a rash during sulfonamide treatment in patients with AIDS) and is, therefore, not of low incidence, be accurately called idiosyncratic? Hence, defining reactions based upon mechanisms, where known, is a preferable means of classification. Where evidence of an immune-mediated response does exist, classification of reactions as immediate-type immune-mediated or delayed-type immune-mediated drug reactions will be used in this review.

Immediate-type immune-mediated (also referred to as Type I hypersensitivity) reactions have been widely studied and are mediated by IgE. As such, a sensitizing exposure to the offending drug must occur prior to elicitation of an IgE response. The sensitizing period may occur during a previous administration of the drug or during the initial course of treatment, followed by elicitation as the drug regimen is continued. It is also important to recognize that sensitization may occur via environmental exposure to agents with antigenic determinants similar to those expressed by a drug or drug-protein complex, which may subsequently elicit an immediate-type immune-mediated reaction upon initial exposure to the drug. For example, Baldo and Pharm (1994) have demonstrated that sera in one patient with an anaphylactic reaction to a commercial hair treatment contained IgE, which reacted with neuromuscular blocking agents; a reaction that was inhibited by a quaternary ammonium derivative in the hair treatment preparation. Moreover, this component in the commercial preparation was able to inhibit the binding of d-tubocurarine to IgE in the sera of three patients who previously experienced life-threatening reactions to a neuromuscular blocker. Thus, failure to identify prior exposure to a drug does not rule out the possibility of an immediate-type immune-mediated drug reaction upon initiation of therapy.

Unlike most antibody isotypes, IgE is primarily located in tissues bound to mast cells. Immediate-type immune-mediated reactions are elicited when an allergen cross-links with preformed IgE on mast cells. This binding results in the degranulation of mast cells, resulting in the release of a variety of toxic mediators (histamine, heparin), cytokines [interleukin (IL)-3, IL-4, IL-5, tumor necrosis factor  (TNF)], lipid mediators (leukotrienes C4 and D4, platelet-activating factor), and enzymes (trypase, chymase, cathepsin G, carboxypeptidase) (Janeway et al., 1999). The descriptor, immediate-type immune-mediated reaction, denotes the rapidity with which the reaction develops, generally occurring in two phases. The immediate phase can occur within seconds of drug exposure and results from the rapid release of toxic mediators, especially histamine, causing an increase in vascular permeability and contraction of smooth muscle. When provoked via systemic administration, this results in a disseminated reaction known as urticaria (or hives), which is manifested as large, red and itchy welts on the skin. When precipitated by local intradermal injection (as occurs in allergy testing or an insect sting), the result is the classic wheal-and-flare reaction. A late-phase reaction generally follows that is mediated by the expression of cytokines and chemokines, resulting in recruitment of a variety of inflammatory cells, including eosinophils. This recruitment causes a more widespread edematous reaction. Persistence of antigen can result in a chronic inflammatory condition.

An immediate-type immune-mediated reaction may be associated with disseminated mast cell activation, resulting in marked changes in vascular permeability, epiglottal swelling, constriction of airways, and vascular collapse. This syndrome is known as anaphylactic shock and represents a life-threatening phenomenon, potentially resulting in death within minutes.

The signals that initiate mast cell degranulation appear to necessitate the formation of antigen-antibody complexes that result in the cross-linking of adjacent IgE molecules (Baldo and Pharm, 1994; Janeway et al., 1999). This would appear to exclude small molecules, such as drugs, from eliciting such reactions. Because clinical experience substantiates that small molecules do in fact elicit such reactions, it was postulated that linkage of the drug with an endogenous protein was necessary to create the molecular size necessary for such cross-linking. Although quaternary ammonium compounds may represent an exception, numerous studies have demonstrated the importance of hapten-protein conjugate formation in the provocation of immediate-type immune-mediated drug toxicity (Didier et al., 1987).

The basis for our current understanding of the role of hapten-protein conjugates in immune-mediated reactions to small molecules is derived largely from studies on the mechanism of these reactions induced by penicillin (Ahlstedt and Kristofferson, 1982). Penicillin is a reactive compound that readily decomposes to form several compounds that have been demonstrated to covalently bind to proteins and represent the primary antigenic determinants of penicillin (Fig. 5). Of the penicillin molecules that covalently bind to protein under physiological conditions, 95% form penicilloyl groups (Baldo, 1999). This forms the basis for the designation of this as a "major" antigenic determinant, rather than inherent immunogenicity. Sera from subjects who have experienced an immediate-type reaction to penicillin exhibit heterogeneity in the IgE antibody response to antigenic determinants, supporting the presence of multiple antigenic determinants (Harle and Baldo, 1990). Conjugation most often occurs via amide linkage to the side chain -amino group on lysine residues, but may also occur via other residues (e.g., tyrosine).

View larger version (20K): [in this window] [in a new window]   Fig. 5.   Structures of the major and minor penicillin antigenic determinants with identified points of attachment to carrier protein. Examples of R groups are illustrated in Fig. 6. Adapted from Baldo and Pharm (1994) with permission.

Differences among penicillin derivatives are based in the side chain, often resulting in differences in reactivity. Antigenic recognition of the common -lactam ring or thiazolidine rings by antibodies in a patient is indicative of a high degree of cross-reactivity among the various penicillin derivatives. In contrast, side chain group recognition will result in considerably less cross-reactivity (Fig. 6), meaning that those derivatives with dissimilar side chains may not evoke an immune response in a patient with a history of immediate-type immune-mediated reaction toward one of the penicillins (Baldo et al., 1995; Baldo, 1999).

View larger version (22K): [in this window] [in a new window]   Fig. 6.   Specific antigenic recognition of methyl-phenyl-isoxazoyl side chain determinants on penicillin by IgE in the sera of a patient who exhibited a reaction to flucloxacillin. Redrawn from Baldo et al. (1995) with permission.

The clinical course of many CDRs suggests they are delayed-type immune-mediated drug reactions. Growing evidence indicates that T-cell recognition of drugs is a critical step in the generation of these responses (Weltzien et al., 1996; Coleman and Blanca, 1998; Pichler et al., 1998). T-cells possess clonally distributed receptors (TCR) that recognize antigen on the cell surface when presented by products of the major histocompatibility complex (MHC) genes. This cell surface-dependent recognition is referred to as MHC-restricted antigen recognition. CD4+ T-helper cells recognize antigen presented by MHC Class II, while those antigens presented by MHC Class I are recognized by CD8+ cytotoxic T-cells.

Evidence for the role of T-cell activation in CDR derives from numerous studies wherein drug-specific T-cell clones have been derived from the peripheral blood of patients with a history of such reactions to amoxicillin, carbamazepine, lidocaine, phenytoin, or SMX (Mauri-Hellweg et al., 1995; Zanni et al., 1996, 1997; Horton et al., 1998). Most of these drug-specific clones express the TCR type, although some clones have been identified that express the TCR type (Zanni et al., 1997; von Greyerz et al., 1999). Hertl et al. (1995) have demonstrated that dermal T-cells isolated from skin lesions in a patient with severe blistering exanthem caused by trimethoprim-SMX (TMP-SMX) proliferated in response to SMX (but not TMP) in the presence of autologous mononuclear cells (used as antigen-presenting cells, APC). Immunohistochemical studies have demonstrated the infiltration of CD4+ and CD8+ T-cells in drug-induced TEN (Leyva et al., 2000).

Because most drugs are of low molecular weight, their recognition as antigens is believed to necessitate the haptenation discussed previously. Few agents are chemically reactive and able to covalently bind with proteins and, thereby, directly act as haptens. Most drugs that evoke a delayed-type immune-mediated reaction are believed to undergo bioactivation to generate haptens that are recognized by sensitized lymphocytes (Merk et al., 1997). Hence, many investigations have focused on the role of bioactivation of causative agents to reactive metabolites, which may in turn bind to critical macromolecules and initiate the immune cascade that eventually presents as a delayed-type immune-mediated CDR. Numerous studies have demonstrated that several sulfonamides, as well as the sulfone dapsone, can undergo biotransformation to form reactive N-hydroxylamine metabolites (Rieder et al., 1988; Coleman et al., 1989; Cribb and Spielberg, 1990a,b, 1992; Cribb et al., 1996b). These studies indicate that the ability to detoxify these reactive metabolites may be an important determinant in the development of CDR. Interestingly, several studies have suggested that the in vitro cytotoxicity of these hydroxylamines toward peripheral blood mononuclear cells can predict predisposition to delayed-type immune-mediated reactions toward sulfonamides (Shear et al., 1986; Rieder et al., 1989; Carr et al., 1993b; Wolkenstein et al., 1995b).

The anticonvulsants phenytoin and carbamazepine (which are also among the highest CDR-producing drugs) can also be metabolized by liver microsomes to reactive metabolites that are cytotoxic and exhibit covalent binding to macromolecules (Spielberg et al., 1981b; Roy and Snodgrass, 1990; Pirmohamed et al., 1992; Furst and Uetrecht, 1993; Castren et al., 1996). It has also been demonstrated that lymphocytes isolated from patients with a history of a CDR to phenytoin or carbamazepine incubated with parent drug and liver microsomes exhibit a higher cell death than lymphocytes from patients without a history of such a reaction (Shear and Spielberg, 1988; Wolkenstein et al., 1995b). This suggests that a defect in detoxifying enzymes may be an important predisposing factor for the development of CDRs to anticonvulsants.

One question that arises from these observations is, how can reactive metabolites generated in the liver survive transit to the sites where delayed-type hypersensitivity is manifested (most commonly, the skin)? It has been proposed that circulating immune cells may catalyze the bioactivation of these agents (Uetrecht, 1992; Furst and Uetrecht, 1993; Uetrecht et al., 1993). Indeed, activated monocytes and neutrophils have been shown to convert numerous CDR-evoking drugs to reactive metabolites (Furst and Uetrecht, 1993, 1995; Uetrecht et al., 1993; Lai et al., 1999). Using aniline as a model compound, Wulferink et al. (2001) have demonstrated that prohaptens incubated with white bone marrow cells form neoantigens that are recognized by T-cells. To date, however, there is no explanation for why reactive metabolites generated in circulating immune cells would display the type of specificity in reaction site that is reported with drugs associated with these reactions. We have, therefore, developed a working hypothesis wherein events occurring in the cutaneous environment are the primary factors initiating these reactions (Reilly et al., 2000).

As shown in Fig. 7, it is proposed that bioactivation occurs in or near keratinocytes. Wolkenstein et al. (1998) have demonstrated that isoforms of cytochrome P450 that form the reactive metabolite of carbamazepine are present in the skin. In addition, we have demonstrated that cultured human keratinocytes are able to bioactivate SMX and dapsone to their respective hydroxylamine metabolites (Reilly et al., 2000). Moreover, the well known photosensitivity of patients receiving numerous drugs that evoke a delayed-type immune-mediated CDR suggests these drugs do reach the cutaneous level after systemic administration (Selvaag, 1997; Vassileva et al., 1998; Epstein, 1999).

View larger version (38K): [in this window] [in a new window]   Fig. 7.   Working hypothesis for the mechanism of cutaneous drug reactions. Adapted from Reilly et al. (2000) with permission. CD80, B7.1 costimulatory molecule.

Long recognized for its importance in the structural integrity of the skin, the epidermal keratinocyte is a key cell in the initiation and propagation of cutaneous immune reactions (Stoof et al., 1994; Nickoloff et al., 1995). Keratinocyte-derived cytokines provide essential signals for the migration of Langerhans cells, which are believed to be the predominant APC in the skin (Cumberbatch et al., 1994, 1997a,b, 1999; Cumberbatch and Kimber, 1995). Modulation of the expression of these keratinocyte-derived cytokines may, therefore, impact antigen presentation and, ultimately, T-cell activation.

As shown in Fig. 7, we hypothesize three avenues that may be important in the initiation of CDR after exposure to reactive metabolites: 1) a direct cytotoxic effect, 2) stimulation of danger signal(s) expression, and/or 3) stimulation of adhesion/costimulatory molecule expression. Importantly, we have demonstrated that incubation of SMX, dapsone, or their respective hydroxylamine metabolites, with normal human keratinocytes results in the formation of drug/metabolite-protein adducts, even in the absence of cytotoxicity (Reilly et al., 2000; W. L. Wurster and C. K. Svensson, unpublished observations). Thus, antigen presentation may occur as the result of cytotoxicity or in its absence. Since keratinocytes have been shown to express human leukocyte antigen (HLA-DR) at inflammatory sites (Barbaud et al., 1997), it is possible that these cells may serve as APC in situations that predispose subjects to CDRs. Moreover, keratinocytes have been demonstrated to express accessory molecules that serve as important "second signals" in the activation of T-cells (Nickoloff et al., 1993, 1995).

Binding of antigen to an antigen-specific receptor (i.e., TCR) is, in itself, insufficient to evoke an immune response. Costimulatory signals are necessary to trigger clonal activation and expansion of T-cells, and in their absence, clonal inhibition and elimination occur (see Figs. 8 and 9) (Goodnow, 1996; van Parijs and Abbas, 1998). The prevailing view has long been that exogenous signals are essential in the control of this response, but growing evidence indicates that endogenous signals may be of equal or greater importance in some immune responses (Ibrahim et al., 1995; Matzinger and Fuchs, 1996; Matzinger, 1998). It has been suggested that damaged or stressed cells release "danger (alarm) signals" that activate local APC, resulting in antigen uptake, migration to the draining lymph nodes, and up-regulation of further stimulatory signals important for the activation of T-cells. Indeed, resident dendritic cells may serve as sentinel cells, responding to signals of cell stress or "unnatural" cell death. The detection of these events may be the primary determinant of the immune system's response to foreign elements. For example, numerous experiments have demonstrated that resident Langerhans cells in the skin acquire dendritic cell properties when exposed to cytokines such as granulocyte-macrophage-colony stimulating factor, TNF- and IL-1 (Lappin et al., 1996; Kimber et al., 1998). Injured cells (e.g., keratinocytes) may serve as the primary source for these cytokines. It has recently been suggested that altered surface expression of heat shock proteins (HSP) may serve as a primary signal of cell stress in response to sublethal cell injury (Ibrahim et al., 1995; Matzinger and Fuchs, 1996; Matzinger, 1998; Todryk et al., 2000). It has also been suggested that T-cells may serve a critical function in the surveillance for damaged cells, which are subsequently eliminated (Boismenu and Havran, 1997, 1998; Salerno and Dieli, 1998). Similar signals have been hypothesized to play a role in the response to this cell type, especially HSP. Interestingly, it has been reported that both HSP27 and HSP70 were expressed in the epidermis during severe drug eruption (Nishioka et al., 1999). Thus, dysregulated cell death/stress resulting from exposure to reactive metabolites may result in important signals that stimulate APC activation, including heat shock proteins and cytokines that have been demonstrated to be important in Langerhans' cell migration.

View larger version (17K): [in this window] [in a new window]   Fig. 8.   Antigen-presenting cell and T-cell interaction. Langerhans' cells in the skin become potent APC due to the expression of critical molecules such as class II MHC. They also increase their expression of accessory molecules, such as ICAM-1, CD58, and CD80/86. These accessory molecules allow T-cells to adhere to APC and also deliver "second signals" to T-cells to amplify the APC-derived "first signal" (i.e., MHC II-hapten complex). Adapted from Gaspari (1993) with permission.

View larger version (25K): [in this window] [in a new window]   Fig. 9.   Potential consequences of antigen presentation by APC. T-cell activation by APC, such as Langerhans' cells in the skin, requires two signals. Antigen presentation on MHC II molecules (which represents signal 1) in the absence of signal 2 results in clonal anergy and tolerance. Presentation of antigen together with signal 2 results in T-cell activation and proliferation, the liberation of lymphokines, and initiation of events that manifest as a CDR. Endogenous signals (i.e., "danger" signals) from stressed cells may induce the expression of CD80/86 on APC, thereby providing the second signal for T-cell activation. Gray oval represents antigen. Adapted from Gaspari (1993) with permission.

T-cell activation by APC necessitates two signals (Fig. 9). The interaction of the antigen presented on major histocompatibility complex and its cognate TCR provides the first signal (Martin and Weltzien, 1994). The second signal arises from the interaction between a costimulatory molecule and its specific receptor on the T-cell surface. Langerhans' cells (the primary APC in the skin) express numerous costimulatory molecules, including CD80 (B7.1) and CD86 (B7.2) (Gaspari, 1993; Nickoloff and Turka, 1994). Importantly, we have recently demonstrated that CD80 may be expressed in normal human keratinocytes exposed to contact irritants and allergens (Wakem et al., 2000). This suggests the possibility that exposure of keratinocytes to reactive metabolites may induce expression of CD80 and allow these cells to provide the critical second signal for T-cell activation.

Intercellular adhesion molecule-1 (ICAM-1; also known as CD54) plays an important role in cell-cell-mediated immune responses, including drawing cells to the local enviroment (Boyd et al., 1988). Common sensitizing agents, such as nickel, p-phenylenediamine, and urushiol (the antigen in poison ivy) have been shown to induce the expression of ICAM-1 in cultured human keratinocytes (Griffiths and Nickoloff, 1989; Picardo et al., 1992; Gueniche et al., 1994a,b). Other investigators have provided evidence that this induction of ICAM-1 is secondary to oxidative stress (Little et al., 1998; Camins et al., 1999). Induction of ICAM-1 upon exposure to reactive metabolites, which may in turn induce oxidative stress, might be expected to enhance lymphocyte infiltration, making it an important step in initiation of these CDRs.

It should be noted that a new model for recognition of small molecules by activated T-cells has recently been proposed that is metabolism-independent and does not require antigen processing (Horton et al., 1998; Pichler et al., 1998; Zanni et al., 1998). It is postulated that drugs may complex with preformed MHC-peptide complexes already expressed on the cell surface. Incubation of SMX or amoxicillin with glutaraldehyde-fixed APC (which are unable to process antigen but can present preprocessed antigen) results in the activation of drug-specific T-cell clones. Because a washing step negated the activation, it appears that the binding occurs in a noncovalent fashion (Zanni et al., 1998). Although these preliminary observations suggest an alternative pathway for T-cell activation, the demonstration that metabolism and processing are not essential does not mean they are not involved in the reactions. It remains to be determined whether or not incubation with reactive metabolites of the compounds would provoke a more profound activation than that seen via incubation with the parent drug. Moreover, the vast majority of evidence indicates that such reactions do not occur in the absence of reactive metabolite formation (Uetrecht, 1999). In addition, if these reactions are mediated by nonreactive SMX interacting with preformed MHC-peptide complexes, it is difficult to understand why more patients receiving the drug do not manifest CDR. Whether or not a similar process can mediate sensitization remains to be explored.

The initial step in any photosensitivity reaction is the absorption of a photon. Thus, only those agents with an absorption spectrum in the range of sunlight are associated with photosensitivity. The two primary forms of drug-induced photosensitivity are phototoxicity and photoallergy. Some drugs, such as several of the fluoroquinolones, may induce photosensitivity by both mechanisms (Tokura, 1998).

1. Phototoxicity. Phototoxic reactions are those in which the photoactivated drug or photodegradation products cause direct cellular damage. The reaction is dependent upon the amount of the compound, level of activating radiation, and the quantity of other chromophores in the skin (e.g., epidermal keratins, melanin, etc.) (Shimoda et al., 2000). For example, using data from the French Pharmacovigilance System, Pierfitte et al. (2000) found a strong correlation between UV irradiation and the reporting rate of phototoxic reactions to the fluoroquinolone, sparfloxacin. In fact, this is perhaps the clearest demonstration of the relationship between an environmental factor and an ADR. Under the right conditions, all individuals exposed to the drug and sunlight in adequate levels would be expected to manifest photosensitivity.

View larger version (18K): [in this window] [in a new window]   Fig. 10.   General scheme for the mechanism of phototoxic reactions.

2. Photoallergy. Drug-induced photoallergy exhibits the characteristics of a delayed-type immune-mediated reaction. These reactions only develop in sensitized individuals and appear to be cell-mediated. The earliest evidence of an immune-mediated photosensitivity pathway arose from studies with 3,3',4',5-tetrachlorosalicylanilide (TCSA), an allergic contact photosensitizing agent (Kochevar and Harber, 1977; Takigawa and Miyachi, 1982; Tokura et al., 1991). With exposure to UVA, TCSA covalently couples to isolated protein and cells (Kochevar and Harber, 1977; Tokura et al., 1991). T-cell-mediated responses to TCSA-haptenated epidermal cells have been demonstrated in vivo in the mouse.

Numerous drug-induced autoimmune diseases are accompanied by cutaneous manifestations. Among these are pemphigus-like skin reactions. A common feature of both idiopathic and drug-induced pemphigus is autoantibodies against one or more keratinocyte cell-surface molecules, including desmoglein I (Koulu et al., 1984; Amagai et al., 1991; Kuechle et al., 1994a). Unlike the case with idiopathic pemphigus, however, the severity of the drug-induced form does not appear to correlate with the level of circulating autoantibodies (O'Loughlin et al., 1978; Kuechle et al., 1994a). The majority of drugs that induce pemphigus-like skin reactions are sulfhydryl-containing compounds, such as penicillamine and captopril. These agents are capable of forming disulfide-linkages to sulfhydryl-containing cysteine residues on a variety of peptides. This ability, however, does not by itself explain the production of autoantibodies. For example, Coleman and associates found that although injection of D-penicillamine resulted in the formation of plasma protein conjugates in vivo, no detectable antibody response was produced in rats; while a conjugate of the drug with keyhole limpet hemocyanin readily induced an immune response (Foster et al., 1987; Coleman et al., 1988). Moreover, drug-induced pemphigus has also been reported with the angiotensin-converting enzyme inhibitor enalapril, which is structurally similar to captopril, but lacks the sulfhydryl moiety believed to play a role in these reactions (see Fig. 11) (Kuechle et al., 1994a). A direct effect of drugs on keratinocytes cannot be ruled out as a possible mechanism for these reactions.

View larger version (11K): [in this window] [in a new window]   Fig. 11.   Structures of captopril and enalapril.

Linear IgA bullous dermatosis is another autoimmune disease that may be associated with drug ingestion. This blistering disorder of the subepidermal region is characterized by the linear deposition of IgA in the basement membrane zone (Kuechle et al., 1994b). Drugs implicated in this disorder include amiodarone, captopril, diclofenac, phenytoin, and, most commonly, vancomycin (Acostamadiedo et al., 1998; Klein and Callen, 2000). Although little is known about the mechanism of this disorder, two potential antigens have been identified; a 97-kDa protein in the lamina lucida and a 285-kDa protein in the lamina lucida and sublamina densa regions (Wojnarowska et al., 1991; Zone et al., 1998). How drug ingestion may expose or provoke an antibody response to critical antigenic determinants on these proteins is unknown. Interestingly, there is evidence that triggering events, such as infection, may be necessary to initiate an immune response in this disorder (Thune et al., 1984; Blickenstaff et al., 1988; Godfrey et al., 1990). Thus, it is possible that drugs do not directly cause this disease, but rather represent one of many possible triggering factors for the idiopathic autoimmune disorder.

	IV. Role of Viral Infection as a Predisposing Factor for Cutaneous Drug Reactions

Top Previous Next References

One of the most perplexing aspects of CDRs is their highly variable nature. In general, only a small fraction of patients who receive a drug will experience a CDR. In addition, some patients who have exhibited a cutaneous reaction that appears reasonably well linked to the suspected drug receive that agent at a later date without experiencing the same reaction (Bonfanti et al., 2000; Carr et al., 1993a; Shafer et al., 1989). Is this because the suspected drug was not the actual cause of the cutaneous reaction? Or could it be that other mitigating factors were necessary for the reaction and those factors were not present upon subsequent administration of the drug? While many investigators have concentrated on identification of innate predisposing factors in patients exhibiting CDRs (e.g., pharmacogenetics), our growing understanding of the multi-factorial nature of many diseases suggests that the presence of transient phenomenon concurrent with drug administration may be critical in determining whether or not a patient exhibits an ADR (Evans, 1982). One factor that appears to significantly increase the risk of CDR is the presence of a viral infection (Haverkos et al., 1991; Levy, 1997).

Pullen et al. (1967) published a study examining the incidence of drug rashes in 184 patients diagnosed with infectious mononucleosis. Of the 121 patients who received one or more antibiotic, 45% exhibited a skin rash, whereas a rash occurred in only 16% of patients not receiving an antibiotic. Rashes were most frequent among patients receiving ampicillin, where 18 of 19 patients receiving the drug experienced an extensive maculopapular, pruritic rash, often accompanied with fever. In contrast, only 2 of 17 patients treated with tetracycline exhibited a rash. Patel (1967) reported on the frequency of skin rash in patients with infectious mononucleosis receiving no antibiotic, receiving ampicillin, or receiving other antibiotics (including penicillin and tetracycline). The frequency of skin rash was 9, 100, and 14%, respectively. Similarly, in a series of patients with cytomegalovirus infection, 4 of 5 patients receiving ampicillin exhibited a rash, while only 2 of 23 patients not receiving ampicillin had skin manifestations (Klemola, 1970). This increased sensitivity to ampicillin in the presence of viral infection was shown to be a transient phenomenon in that only 2 of 20 patients experiencing an ampicillin-induced rash during infectious mononucleosis exhibited a recurrent rash when receiving the drug after resolution of the infection (Nazareth et al., 1972).

The incidence of CDRs among patients infected with the human immunodeficiency virus (HIV) is also substantially increased compared with that found in presumed HIV-negative subjects (Coopman and Stern, 1991; Coopman et al., 1993). For example, the Boston Collaborative Drug Surveillance Study found that the incidence of CDR to TMP-SMX in hospitalized patients was 3.3% (Bigby et al., 1986). The frequency of CDRs to TMP-SMX in HIV-infected subjects is reported to be as high as 40 to 60% (Medina et al., 1990; Blum et al., 1992; Pertel and Hirschtick, 1994; Roudier et al., 1994). This increased frequency of CDR does not appear to be secondary to the effects of Pneumocystis carinii pneumonia (PCP) infection (the most common indication for the drug in HIV-infected subjects) or the higher doses used in the treatment of PCP. Kovacs et al. (1984) found that CDRs to TMP-SMX in patients being treated for PCP occurred in 10 of 34 HIV-infected patients, but in 0 of 15 patients with other immunosuppressive diseases. In other studies where TMP-SMX has been used in immunosuppressed patients (presumed to be HIV-negative), CDRs have been extremely rare (Winston et al., 1980; Sattler and Remington, 1981). Similarly, the incidence of CDR during treatment with dapsone appears to be higher in HIV-infected subjects (Medina et al., 1990; Blum et al., 1992; Pertel and Hirschtick, 1994). Thiacetazone causes a CDR in 20 to 29% of HIV-infected patients with tuberculosis, but only in 1 to 7% of HIV-negative tuberculosis patients (Nunn et al., 1991; Watkins et al., 1996). Other drugs have been noted to cause an unusually high frequency of CDRs in HIV-infected patients (see Table 2), although whether this represents an "increased" incidence is unclear, since the only indication for the agents is HIV-infection (Ho et al., 1998; Anon, 2000; Centers for Disease Control, 2001).

By what mechanism(s) do viral infections increase the predisposition to the development of CDRs? The working hypothesis that we have previously proposed provides a framework for understanding the potential role of alterations in key immune mediators in predisposing patients to these reactions (Fig. 7) (Reilly et al., 2000). It is well known that immune responses in the skin are highly regulated by cytokines and other inflammatory mediators (Kimber et al., 1995; Kimber, 1996; Kimber and Dearman, 1996). Viral and bacterial infections can also result in the release of a variety of cytokines that may up-regulate the expression of key immune-mediating molecules in keratinocytes and Langerhans' cells. For instance, it has also been demonstrated that cyclooxygenase 2 (COX-2) is up-regulated in keratinocytes in an inflammatory response (Lotfin and Eling, 1996; Maldve and Fischer, 1996; Buckman et al., 1998). If COX-2 contributes to the bioactivation of arylamines and other compounds as some studies have suggested (Liu and Levy, 1998; Goebel et al., 1999), an inflammatory response would increase the formation of reactive metabolites in the keratinocyte. Other xenobiotic-metabolizing enzymes have been shown to be either up- or down-regulated in other cells upon exposure to pro-inflammatory cytokines (Morgan, 1997, 2001). In addition, we have shown that glutathione content is a critical factor in determining the susceptibility of keratinocytes to the hydroxylamine metabolite of SMX (Reilly et al., 2000). This observation is significant in that the pro-inflammatory cytokines TNF- and IL-1 induce a variety of biochemical responses in cells, including decreases in glutathione (Singh et al., 1998). Interestingly, TNF- is substantially increased in the epidermis of patients with severe CDR (Barbaud et al., 1997). Thus, by their multiple immunological and biochemical effects, viral infections may "prime" the cutaneous environment for a response to agents that result in a CDR.

	V. Cutaneous Drug Reactions Associated with Selected Drugs

Top Previous Next References

The incidence of CDR in patients receiving TMP-SMX is higher than for any other drug (Naldi et al., 1999). As the most widely used sulfonamide preparation, it is not surprising that most reported sulfonamide-related CDRs involve this combination. Although it is believed that the SMX component is responsible for most reactions, CDR in patients receiving trimethoprim monotherapy have also been reported (Das et al., 1988; Nwokolo et al., 1988). As described previously, the frequency of TMP-SMX-associated CDR is substantially increased in HIV-infected subjects (Medina et al., 1990; Blum et al., 1992; Pertel and Hirschtick, 1994; Roudier et al., 1994).

Sulfonamides have been reported to cause a wide range of skin reactions, including erythema multiforme, erythema nodosum, photosensitivity, TEN, and urticarial rashes (Cribb et al., 1996a). The mechanism and time course of these reactions varies considerably. Episodes of urticarial rash, generally not accompanied by fever, can occur within the first 3 days of therapy with a sulfonamide. Although anaphylaxis may not accompany the initial CDR, re-exposure can provoke this life-threatening reaction. These reactions appear to be immediate-type immune-mediated reactions that are associated with the presence of IgE antibodies toward the sulfonamide (Carrington et al., 1987; Harle et al., 1988; Gruchalla and Sullivan, 1991). Binding specificity studies suggest the 5-methyl-3-isoxazolyl group on SMX is a key component in antibody recognition (Harle et al., 1988). In vitro cross-reactivity with a variety of sulfonamides suggests such reactions may occur across a spectrum of sulfonamides in a given patient. The response to skin testing is variable and does not appear to identify a significant portion of patients with a positive history of urticarial CDR to sulfonamides (Leftwich, 1944; Gruchalla and Sullivan, 1991).

Sulfonamides also induce a delayed-type reaction that appears to be immune-mediated. These reactions are generally manifested 7 to 14 days after the initiation of therapy with the sulfonamide and most commonly present as fever, together with a morbilliform or maculopapular, non-urticarial skin rash. Patients may also progress to SJS or TEN. The incidence of these severe cutaneous reactions with sulfonamides is estimated to be between 1:1000 and 1:100,000 (Chan et al., 1990; Roujeau and Stern, 1994). In addition to CDRs, sulfonamides have been associated with a multiorgan syndrome, most frequently involving fever, rash, eosinophilia, and hepatotoxicity, which develops after a week or more of therapy (Dujovne et al., 1967; Berg and Daniel, 1987; Rieder et al., 1989). The incidence of this multiorgan syndrome appears to be less than 0.1% (Cribb et al., 1996a).

The ability to demonstrate drug-specific activated T-cell clones in patients exhibiting these delayed-type reactions supports the contention that they are immune-mediated (Warrington et al., 1983; Kalish et al., 1994; Hertl et al., 1995; Mauri-Hellweg et al., 1995; Schnyder et al., 1997, 1998; von Greyerz et al., 1999). Such clones are, however, at low frequency and not detectable in all patients with these delayed-type reactions.

Numerous lines of evidence suggest the need for bioactivation of the sulfonamide prior to initiation of a CDR. Figure 12 illustrates the proposed bioactivation-dependent mechanism for SMX-induced delayed-type immune-mediated reactions. The first evidence of the importance of metabolism as a predisposing factor for these reactions was provided by Shear et al. (1986), who found that six of six children who experienced serious reactions (CDR plus systemic organ involvement) to a sulfonamide exhibited the slow acetylator phenotype. In contrast, a cohort of control subjects exhibited a frequency of slow acetylators of 56%. Moreover, lymphocytes from these patients exposed in vitro to several sulfonamides plus murine liver microsomes exhibited a higher percentage of cell death than lymphocytes from control subjects (Shear et al., 1986). These data suggest that subjects with a delayed-type immune-mediated reaction to sulfonamides may have a defect that reduces their ability to detoxify reactive sulfonamide metabolites. It also suggests that patients with the slow acetylator phenotype may be at greater risk of such reactions because they are presumed to bioactivate a larger fraction of the administered dose via cytochromes P450. Subsequent studies have confirmed this latter observation in patients experiencing a sulfonamide-induced delayed-type immune-mediated reaction (Rieder et al., 1988; Wolkenstein et al., 1995b). Moreover, it has been demonstrated that humans metabolize SMX to a hydroxylamine metabolite that exerts a cytotoxic effect on peripheral blo