Histopathology of drug eruptions – general criteria, common patterns, and differential diagnosis

Drug eruptions are among the most common inflammatory diseases of the skin and also among those biopsied most often. Yet, the value of histopathologic examination of drug eruptions has often been disputed. One reason is that the spectrum of histopathologic changes in drug eruptions is broad. Nevertheless, each histopathologic pattern assumed by drug eruptions has a limited number of differential diagnoses, and numerous criteria and clues are available to distinguish drug eruptions from other diseases associated with those patterns. By recognition of common patterns, consideration of differential diagnoses, and attention to distinct clues, a histopathologic diagnosis of drug eruption can usually be made with confidence.

by distinctive changes, such as ballooning and multinucleated keratocytes in measles and keratocytes with steel-grey nuclei with margination of nucleoplasm in infections by herpesviruses [11][12]. Often, however, there are no such distinguishing features. Common patterns of viral exanthemata include a superficial perivascular infiltrate of lymphocytes without associated epidermal changes, a superficial vacuolar interface dermatitis, sometimes associated with eosinophils and neutrophils, a lichenoid dermatitis, and a mild spongiotic dermatitis. All those patterns may also be encountered in drug eruptions [10].
Despite those limitations, drug eruptions can usually be diagnosed with confidence on the basis of histopathologic changes alone. It is common practice in laboratories of dermatopathology to examine sections of biopsy specimens before obtaining any clinical information. If this is done, and the presumptive diagnosis of a drug eruption is rendered, it is our experience that the latter is usually corroborated by the clinical diagnosis of the referring physician. All histopathologic diagnoses must be submitted to critical review in the context of additional information, such as clinical findings, clinical history, and laboratory data. This does not distinguish drug eruptions from any other skin disease, and the reliability of histopathologic diagnosis of a drug eruption is not smaller than that of diseases for which biopsy is recommended without reservation, be it lichen planus, lupus erythematosus, or granuloma annulare.
In the following, we wish to discuss criteria that aid in recognition of drug eruptions in general, describe common patterns of drug eruptions, and discuss the differential diagnosis of those patterns. The statements made are based on personal experience of many years, on a review of the literature, and on systematic analysis of histopathologic findings in 60 cases of maculopapular drug eruption in which the eliciting drugs were known and the eruption cleared following cessation of them. In that study, cases diagnosed clinically In 1997, Ackerman emphasized that "drugs can elicit any of the nine basic patterns of inflammatory diseases in the skin, and none of those patterns is specific for a drug eruption.
There is but one exception, to date, to the precept that drug eruptions cannot be diagnosed with specificity through the microscope, namely, fixed drug eruption" [1].
In more than a decade following that sobering assessment of the import of histopathological analysis for the diagnosis of drug eruptions, only little progress has been made. For some differential diagnoses, criteria have been set forth to facilitate distinction of drug eruptions from other inflammatory skin diseases, e.g., lichenoid drug eruption from lichen planus, psoriasiform drug eruption from psoriasis vulgaris, and granulomatous drug eruption from granuloma annulare.
In a recent review of histopathologic patterns of cutaneous drug eruptions, one finding indicative of drug eruptions in general has been noted, namely, combination of different patterns in a single specimen. Another finding mentioned as "a diagnostic clue" to drug eruptions was presence of eosinophils, but the authors emphasized that "one must be cautious not to consider them the panacea of histologic diagnosis for a drug reaction as their presence does not make a drug reaction the correct diagnosis. Conversely, the absence of eosinophils does not rule out a drug eruption. In other words, they may or may not be present in these reactions" [2].
The vagueness of histopathologic descriptions of drug eruptions, and the caution exercised in interpretation of them, has created the impression that biopsy of drug eruptions has little value. Current textbooks of dermatology emphasize that, in reactions to drugs, "the histological changes are no more distinctive than are the clinical features," the only exceptions being "vegetating iododermas and bromodermas, certain lichenoid eruptions and fixed drug eruptions," in which "the histological changes … are not pathognomonic, but are sufficiently characteristic to be of importance in differential diagnosis" [3]. No such importance is attributed to histopathologic study of morbilliform drug eruptions, which have been estimated to account for 95% of all drug eruptions [4]. The latter are said to show only "non-specific lymphohistiocytic infiltrates in perivascular arrangement.
For that reason, histopathologic examination can contribute only little to diagnosis and differential diagnosis" [5]. As a consequence, it has been stated unequivocally that "biopsy of morbilliform eruptions is not recommended" [6].
In our view, those conclusions are wrong and potentially harmful, as they may lead to incorrect diagnoses and mismanagement of patients. It is true that histopathologic diagnosis of drug eruptions may be difficult, may remain equivocal, and require clinico-pathologic correlation, but this is true for all diseases. Compared to other inflammatory diseases of the skin, histopathologic diagnosis of drug eruptions is impeded by the fact that drugs may not only cause eruptions as drug-induced erythema multiforme, Stevens-Johnson syndrome, toxic epidermal necrolysis, acute generalized exanthematous pustulosis, as well as lichenoid and bullous drug eruptions were excluded. results of that study are being published separately [13]. For this review, we re-examined biopsy specimens of 300 consecutive cases seen in our laboratory in which the diagnosis of a drug eruption of any kind was given both, histopathologically and clinically, but in which data concerning the eliciting drug and follow-up were available for only a minority of patients. The purpose of that endeavor was to assess the relative frequency of different histopathologic patterns of drug eruptions in the routine material of a laboratory of dermatopathology. Findings of that survey are summarized in Table 1.

general criteria
Several findings are typical of drug eruptions in general. Some of them may appear banal. Nevertheless, when encountered in association with a particular pattern, they may help to rule out other diseases associated with that pattern. These findings include:

signs of acuteness
Drug eruptions, as their name indicates, are usually acute, eruptive diseases that appear suddenly and progress rapidly in both, extension and intensity. As a consequence, they are usually biopsied early in their course. Histopathologic evidence of an eruptive disease biopsied early in its course includes • a normal basket-woven cornified layer despite spongiosis or hydrops of keratocytes in the basal or spinous zone (the reason being that the interval of time between onset of the eruption and biopsy of it is too small to permit alterations in the lower epidermis to affect to stratum corneum), • edema of the papillary dermis, • angiectases of capillaries and venules in the superficial dermis, • many neutrophils in the lumina of dilated venules, • extravasation of erythrocytes.

Vacuolar interface dermatitis
The most common histopathologic pattern of drug eruptions is a vacuolar interface dermatitis. The extent of interface changes varies greatly, from extensive vacuolar alteration at the dermo-epidermal junction and many necrotic keratocytes at all levels of the epidermis, as in most cases of fixed drug eruption and toxic epidermal necrolysis, to focal and very subtle changes. The latter may not be apparent immediately.
When a drug eruption is suspected, it is worthwhile to screen all sections of the biopsy specimen for evidence of a subtle vacuolar interface dermatitis.

Presence of neutrophils and eosinophils
Drug eruptions, like many other inflammatory diseases, are often associated with an infiltrate of eosinophils and/or neutrophils. In a recent study of morbilliform drug eruptions, eosinophils were found in 50% and neutrophils in 36% of cases [18]. In our study of maculopapular drug eruptions in which the eliciting agents were known, the numbers were higher, namely, 60% for eosinophils and 50% for neutrophils. In brief, eosinophils and/or neutrophils are present in the majority of drug eruptions. Eosinophils are more common, but because they are seen in such a wide variety of diseases, they are less distinctive for drug eruptions. An infiltrate of neutrophils is rarer but of greater diagnostic import.

several histopathologic patterns in a biopsy specimen
Drug eruptions may present themselves with different histopathologic patterns. Each of those patterns may be caused by a variety of inflammatory skin diseases. A combination of two or more patterns in a single biopsy specimen, however, favors a drug eruption.
As mentioned above, the most common pattern of drug eruptions is an interface dermatitis which is usually also seen when patterns are combined. regardless of the pattern of inflammation, ranging from superficial perivascular to superficial and deep perivascular and interstitial, from spongiotic to granulomatous, and from subcorneal pustular to subepidermal bullous, presence of tiny foci of vacuolar changes at the junction, sometimes associated with but a few necrotic keratocytes, should raise suspicion of a drug eruption (Figure 2 a, b).

several discrete foci of inflammation in a biopsy specimen
Most maculopapular drug eruptions are generalized eruptions in which the degree of inflammation varies. Individual lesions tend to be circumscribed poorly and blend into one another. Accordingly, there may be two or more discrete foci of inflammation in a punch biopsy specimen, separated from one another by areas in which the inflammation is less pronounced. This may help to distinguish drug eruptions from diseases with distinct papules, such as lichen planus and pityriasis lichenoides.

other clues to diagnosis of a drug eruption
Drug eruptions usually affect trunk and extremities. Palms and soles are involved only rarely, and if they are, there are usually also lesions at other sites that are selected for biopsy.
As a consequence, drug eruptions, with the exception of fixed drug eruption, are biopsied rarely on palms and soles.
The same is true for the face and scalp. Hence, when one sees a biopsy specimen from face, scalp, or palmar and plantar surfaces, a drug eruption is unlikely.
Drug eruptions are most common in elderly patients. chromatin. Nuclei may also be hyperchromatic. In contrast to epithelial neoplasms, atypical keratocytes are usually confined to discrete foci and are not crowded together closely. They have been described especially in reactions to chemotherapeutic drugs [19][20]. However, they may be seen in response to a wide variety of drugs and seem to be related to interface changes, since they are also encountered episodically in other interface dermatitides, such as lichen sclerosus and lupus erythematosus. In brief, atypical keratocytes are neither a sensitive nor a specific finding. Nevertheless, because they are more common in drug eruptions than in other inflammatory skin diseases, they may serve as a clue to histopathologic diagnosis of a drug eruption ( Figure   3 a, b).

Common patterns and differential diagnoses lymphocytic dermatitis without epidermal changes
This is the least distinctive pattern of a drug eruption. It is not very common, accounting for only 12 of 300 (4%) consecutive cases examined. Often subtle vacuolar changes at the junction or slight spongiosis in the lower half of the spinous zone can be detected in sections of what, at first blush, seems to be a perivascular lymphocytic dermatitis without epidermal changes. In other cases, scrutiny reveals eosinophils and/or neutrophils in addition to lymphocytes.
When neither subtle vacuolar changes at the junction, spongiosis, nor eosinophils and neutrophils are present, the differential diagnosis includes a wide variety of diseases. A sparse superficial perivascular infiltrate of lymphocytes is physiologic and may be seen in clinically normal skin as well as in the earliest stage of diseases that, at a later stage, are associated with distinctive histopathologic findings. Hence, if the infiltrate is very sparse, a specific diagnosis, and even a meaningful differential diagnosis, cannot be rendered.
A relatively dense perivascular infiltrate of lymphocytes without associated epidermal changes, however, is not physiologic and excludes diseases that, given the degree of inflammation, should also sport additional findings. In that instance, the differential diagnosis includes Schamberg's disease, secondary syphilis, erythema chronicum migrans, polymorphous light eruption, pernio, lupus erythematosus tumidus (including Jessner's lymphocytic infiltration and reticular erythematous mucinosis), viral exanthemata, and drug eruptions.
Drug eruptions with a wholly lymphocytic infiltrate usually affect only the superficial dermis. In the 300 consecutive cases diagnosed clinically as a drug eruption, we encountered only two with a superficial and deep wholly lymphocytic infiltrate. By contrast, the infiltrates in infections by borrelia, polymorphous light eruption, pernio, and tumid lupus ery-thematosus usually affect the entire dermis. Other findings may also be helpful to rule out differential diagnoses, such as mucin in the interstitial dermis in tumid lupus erythematosus, some plasma cells in secondary syphilis, and erythrocytes in dermal papillae and epidermis in Schamberg's disease. Drug eruptions are also associated commonly with extravasation of erythrocytes and may be indistinguishable from Schamberg's disease [2]. In hemorrhagic drug eruptions, however, erythrocytes are mostly seen around venules of the superficial plexus rather than in discrete collections in dermal papillae. When consisting of lymphocytes only, the infiltrate in drug eruptions tends to be restricted to perivascular areas with only little involvement of the interstitium. This helps to distinguish drug eruptions from infections by borrelia which are usually associated with many lymphocytes in the interstitial dermis.

superficial and deep perivascular and interstitial dermatitis with eosinophils and neutrophils
This pattern, in the absence of significant epidermal changes, was found in 12 of 300 consecutive cases of drug eruptions (4%). If the infiltrate is sparse, the most important differential diagnosis is urticaria. In such cases, the epidermis and papillary dermis should be screened for subtle alterations that may be visible only in step sections, such as slight focal spongiosis or interface changes, slight subepidermal fibrosis or some melanophages. Any of those findings militates against urticaria. The same is true for perivascular accentuation of the infiltrate. Other clues to an urticarial drug eruption are pronounced edema of the papillary dermis and more neutrophils than normally seen in urticaria [2]. However, even in cases in which none of those clues to a drug eruption are encountered, causation of urticaria by a drug cannot be ruled out.
If the infiltrate is denser, the most important differential diagnosis is a response to an arthropod assault.

Vacuolar interface dermatitis
Vacuolar interface dermatitis is the most common pattern of drug eruptions. In a recent study of morbilliform drug eruptions, interface changes were described in 53% of cases [18].
In our study of maculopapular drug eruptions, subtle vacuolar changes at the dermo-epidermal junction were noticed in 58 of 60 cases (97%), and of the 300 consecutive cases,

a) severe vacuolar interface dermatitis
This pattern was encountered in 38 of 300 cases (13%), 13 of which were diagnosed clinically as fixed drug eruption.
The latter did not differ substantially from other cases of this group. In 26 cases, numerous eosinophils and neutrophils were present in the infiltrate, and in 10 cases, at least some eosinophils and/or neutrophils could be detected. Only four cases were associated with a wholly lymphocytic infiltrate.
Areas of confluent epidermal necrosis were observed in nine cases, including four cases of fixed drug eruption. In 10 cases (including five of fixed drug eruption), the infiltrate extended into the lower half of the dermis.
The differential diagnosis of these cases includes postherpetic erythema multiforme. The epidermal changes are indistinguishable. In general, the infiltrate in post-herpetic erythema multiforme is more perivascular and restricted to the superficial dermis, but involvement of the interstitium and the lower dermis may occur. In the vast majority of cases of post-herpetic erythema multiforme, the infiltrate is wholly lymphocytic. In the literature, eosinophils have been reported in erythema multiforme but, with rare exceptions [21], no clear distinction was made between post-herpetic and druginduced cases [22][23]. When those cases were distinguished, eosinophils were found to be more common in drug-induced erythema multiforme [24][25]. This corresponds to our own experience. For the purpose of this study, we re-examined biopsy specimens of five patients with recurrent post-herpetic erythema multiforme and four patients with erythema multiforme who were younger than 20 years. In three cases of post-herpetic erythema multiforme, few neutrophils were spotted in the papillary dermis, and in one case, a single eosinophil was found. This differs markedly from the high frequency and often high number of eosinophils and neutrophils in drug eruptions with severe vacuolar interface changes (Figure 4 a, b).
Another clue to drug etiology that has been reported in erythema multiforme is acrosyringeal concentration of necrotic keratocytes, a phenomenon that may be related to concentration of drugs in sweat and to direct toxic effects on eccrine ductal epithelium [25]. We found an accumula-

b) mild vacuolar interface dermatitis
A mild vacuolar interface dermatitis with only subtle vacuolar changes at the dermo-epidermal junction and few, if any, necrotic keratocytes is the most common pattern of drug eruptions. In our study of 300 consecutive drug eruptions, it was observed in 83 cases (28%). As mentioned previously, the constellation of mild vacuolar interface changes and a sparse superficial perivascular and interstitial infiltrate of lymphocytes, eosinophils, and neutrophils is virtually diagnostic of a drug eruption (Figure 5 a, b).
The differential diagnosis of drug eruptions with mild vacuolar interface changes includes diseases normally associated with a more pronounced interface dermatitis, such as lupus erythematosus and acute graft-versus-host disease, but also diseases that are never associated with severe interface changes, including viral exanthemata and some autoimmune bullous diseases, especially the urticarial stage of bullous pemphigoid. Because the latter may also be associated with a superficial perivascular and interstitial infiltrate of eosinophils and neutrophils, distinction of it from a drug eruption may be particularly challenging. A clue to diagnosis of bullous pemphigoid is presence, and sometimes cluster-  All those features are also seen in the patch stage of mycosis fungoides. Moreover, subtle vacuolar changes at the dermo-epidermal junction, lymphocytes with largish nuclei, and eosinophils may be seen in both diseases. In the literature, drug eruptions mimicking mycosis fungoides have been reported especially following intake of carbamazepin and phenytoin [14][15][16][17], but other compounds have also been implicated [37][38][39][40][41]. Because drug eruptions may also simulate mycosis fungoides clinically, differentiation of those diseases is all the more challenging. It has been claimed that drug-induced pseudolymphomas "cannot be differentiated from true lymphomas through clinical, pathological or molecular findings," the only way of differentiation being "resolution of the lesions after the medication involved is suspended" [17].

spongiotic dermatitis
Drug eruptions commonly present themselves as a spongiotic dermatitis. We found a spongiotic dermatitis in 62 of 300 consecutive drug eruptions (21%), and some spongiosis was also present in many other cases in which it was not the predominant pattern. In our study of maculopapular drug eruptions in which the eliciting agents were known, 58 of 60 cases (97%) were associated with at least subtle spongiosis [11]. Most commonly, spongiosis is mild and confined to the lower half of the epidermis. Spongiotic vesicles were seen in less than half of the cases classified as spongiotic dermatitis.
Those vesicles were usually small and confined to one or two  The differential diagnosis of pustular drug eruptions includes pustular psoriasis, deficiency disorders such as necrolytic migratory erythema and acrodermatitis enteropathica, and pemphigus, especially IgA pemphigus. In pemphigus, the infiltrate tends to be relatively evenly distributed.
In the dermis, it is usually restricted to the upper half and does not show significant perivascular accentuation. In the epidermis, neutrophils may be dispersed evenly across a broad front in concert with scant spongiosis [42]. By contrast, the infiltrate in drug eruptions is often accentuated around blood vessels and may be deep as well as superficial.
In the epidermis, neutrophils are not scattered broadly but usually aggregated in discrete foci. Evidently, signs of acantholysis favor pemphigus and militate against a drug eruption, although some acantholytic cells may also be found in pustules of drug eruptions. In cases of doubt, this differential diagnoses can be resolved easily by immunofluorescence studies.
Intra-or subcorneal abscesses in deficiency disorders are usually elongated rather than discrete, as in most cases of pustular drug eruptions. When drug eruptions are associated with elongated abscesses, the infiltrate is usually very dense   [44]. Moreover, spongiosis in pustular drug eruptions has been claimed to be "usually mild, in contrast to that seen in pustular psoriasis." [2] subepidermal bullous dermatitis Autoimmune subepidermal bullous diseases may be induced by drugs, a phenomenon especially common in linear IgA dermatosis. Subepidermal blisters in drug eruptions, however, may also result from an interface dermatitis and, rarely, from massive edema in the papillary dermis. We observed subepidermal blisters in six of 300 consecutive drug eruptions (2%), all of which showed signs of interface dermatitis (Figure 10 a, b). In four of those cases, a clinical differential diagnosis of drug eruption versus bullous pemphigoid was given, and the latter diagnosis was excluded by failure to detect autoantibodies in ELISA and/or immunofluorescence studies.
Histopathologic differentiation between bullous pemphigoid and bullous drug eruptions may be difficult because both diseases, in addition to subepidermal blisters, may show a perivascular and interstitial infiltrate with many eosinophils and some neutrophils in the superficial and mid dermis. Necrotic keratocytes may also be seen in both diseases.
In bullous pemphigoid, however, the latter are restricted to the roof of the blister. Necrotic keratocytes at the edge of the blister, where the epidermis has not yet detached from the dermis, strongly favor a drug eruption. The same is true for other signs of interface dermatitis, including prominent vacuolar alteration at the junction and melanophages in the papillary dermis. Neutrophils are less common in bullous pemphigoid and, when present, usually sparse (Figure 8 a, b).

granulomatous dermatitis
Drug eruptions may be associated with granulomatous inflammation. We observed granulomas in 12 of 300 consecutive drug eruptions (4%). Two patterns of granulomatous inflammation could be distinguished. In five cases, there were one or few small, round to oval, sharply circumscribed granulomas in the upper dermis. In three of those cases, at least one granuloma was situated in close proximity to an eccrine duct, suggesting damage to the duct and leakage of sweat as a possible cause of granulomas. All five cases were associated with epidermal changes, either focal spongiosis (two cases), or foci of interface dermatitis (two cases), or both (one case). The associated epidermal changes distinguished those drug eruptions from the most important differential diagnosis, sarcoidosis (Figures 2 a, b). Another clue to diagnosis of a drug eruption observed in two cases were neutrophils in the lumina of venules, a finding hardly ever observed in sarcoidosis.
The second pattern of granulomatous dermatitis was scatter of histiocytes among collagen bundles in one or more poorly circumscribed areas in the superficial and/or deep dermis. There also was a perivascular lymphocytic infiltrate.
Those changes resembled the interstitial type of granuloma annulare. Of seven cases with that pattern, two were indistinguishable from granuloma annulare. In both, a drug eruption could be diagnosed with confidence because of onset of lesions following administration of a new drug (captopril and allopurinol, respectively) and gradual resolution after cessation of it. In those two cases, numerous eosinophils were present, but the latter may also be seen in granuloma annulare.
The five other cases could be distinguished from granuloma annulare because of associated epidermal changes, namely, interface changes in four and spongiosis in one of them. Subtle signs of an interface dermatitis have been described as a histopathologic clue to diagnosis of a granuloma annularelike drug eruption [45]. In five of our seven cases of granuloma annulare-like drug eruption, eosinophils and neutrophils were sparse or absent. A clue to diagnosis of a drug eruption present in four of seven granuloma annulare-like lesions was presence of neutrophils in the lumina of venules.

leukocytoclastic vasculitis
In our study of 300 consecutive drug eruptions, two cases showed signs of leukocytoclastic vasculitis. In both, a clinical diagnosis of drug eruption had been given because of onset of lesions shortly after administration of a new drug. In one of those cases, the same type of eruption had occurred once before following administration of the same drug (azithromycin). Both cases showed stereotypic features of leukocytoclastic vasculitis, namely, fibrin in the walls of venules, extravasation of erythrocytes, and an inflammatory infiltrate composed of lymphocytes, neutrophils, and eosinophils in concert with nuclear dust. In both cases, there were more eosinophils than normally seen in leukocytoclastic vasculitis, including focal clusters of eosinophils. This is in concurrence with a recent study in which a significantly higher number of eosinophils was found in drug-induced than in non-druginduced cases of leukocytoclastic vasculitis. In that study, the course of drug-induced cases was found to be less severe, with lower incidence of extra-cutaneous involvement and faster resolution [46]. Although presence of many eosinophils does not exclude other causes of leukocytoclastic vasculitis, it may serve as a clue to causation by a drug.

Discussion
Adverse cutaneous reactions to drugs may occur in many different forms. So divergent are the patterns of drug eruptions that they cannot be considered variants of a single pathologic process. Evidently, the cytokines involved in eruptions presenting as a pustular, spongiotic, or severe interface dermatitis must be very different from one another. When several biopsies are taken from the same patient, they usually show the same predominant pattern, although associated findings, such as focal spongiosis in a vacuolar interface dermatitis, may be seen in only one of two biopsy specimens. Moreover, patients with recurrent drug eruptions usually show always the same type of response.
And yet, there is some overlap. Signs of interface dermatitis, for example, are extremely common in drug eruptions.
They are mostly mild and most often seen in maculopapular drug eruptions, but even in the latter, they may be pro- to attach individual cases to one of the categories of patterns.
This, however, is not only unavoidable, but irrelevant for the purpose of distinguishing drug-induced cutaneous eruptions from those not induced by a drug. For that purpose, it is helpful to consider the differential diagnosis of a given pattern and findings that allow drug eruptions to be recognized in that particular context. The categories of patterns dis-