Etiology of Pruritic Papular Eruption With HIV Infection in Uganda FREE

Pruritic papular eruption (PPE) is a substantial cause of human immunodeficiency virus (HIV)–related morbidity in sub-Saharan Africa.^1- 7 Despite 2 decades of reports describing this condition, its etiology has not been elucidated. Empirical therapies provide only minimal relief,⁵ and the severe pruritus and subsequent scarred excoriations subject patients to HIV-related stigma in their communities.¹

Reports of PPE emerged early in the course of the HIV epidemic. Beginning in 1983, studies in the Democratic Republic of the Congo,² Mali,³ Zambia,⁴ Tanzania,⁵ Nigeria,⁶ Togo,⁷ and other African countries described an extremely pruritic, diffuse skin eruption occurring in HIV-infected patients. The typical primary lesion is a firm, discrete, erythematous, urticarial papule.⁸ The concentration of the lesions is highest on the extremities, but the trunk and face are also involved in half of the patients.⁸ Most patients scratch the lesions because of the severe pruritus, leading to excoriated papules, marked postinflammatory hyperpigmentation and, eventually, scarred nodules.⁵ The persistent pruritus is usually refractory to topical steroids and oral antihistamines.⁵ Cases have been reported with similar frequencies in infected men, women, and children.² While numerous cases have also been described in Haiti,⁹ Brazil,¹⁰ and Thailand,¹¹ the eruption has rarely been reported in Europe and North America.²

The prevalence of PPE in Africans and Haitians with HIV varies from 12% to 46%, depending on the geographic area.^2,4,8- 9 It is uncommon in immunocompetent patients,^2,8 with a reported positive predictive value of 82% to 87% for HIV infection.^2,12 More than half of HIV-infected patients in some countries (including Uganda and the Democratic Republic of the Congo) may report the eruption as their initial disease manifestation.^2,9 The incidence of the rash increases with decreasing CD4 cell counts,¹¹ but it occurs even in early disease.² In fact, given the prevalence and positive predictive value of the rash, it has played a role in diagnosing HIV infection in countries in which serologic testing is not available or affordable.³

There have been scattered, inconsistent reports of the pathologic findings from biopsy specimens^2,4; in these reports, most of the specimens showed a mild to moderate dermal perivascular and periadnexal mononuclear infiltrate with variable intermixed eosinophils.² Special stains for organisms have been negative, and all efforts to implicate an opportunistic infection have been unsuccessful.^2,8 Lower levels of interleukin 2 (IL-2) and γ-interferon in HIV-infected patients with the eruption compared with other HIV-infected patients have led some to argue for a role for immune dysregulation.¹⁰ An increase in local and peripheral eosinophilia as well as an increase in IgE levels has been reported, suggesting hypersensitivity to arthropod bites.¹³ These biopsy and laboratory studies, however, have been far too limited in scope to provide any consensus as to the etiology of PPE.

A comprehensive review of PPE published in 1993 suggested that it “most likely represents a spectrum of diseases, including HIV-associated eosinophilic folliculitis, that have similar clinical and histologic findings.”⁸ Known dermatologic diseases with related clinical presentations include staphylococcal folliculitis, Demodex folliculitis, drug eruption, exuberant reactions to arthropod bites, and photodermatitis. Crusted scabies, secondary syphilis, onchodermatitis, and papulonecrotic tuberculids can also resemble PPE.

We studied Ugandan HIV-infected patients with PPE, performing biopsies on early lesions to gain insight into the etiology of the disease.

We performed a cross-sectional, descriptive, biopsy study of HIV-infected patients with active clinical PPE enrolled as a consecutive convenience sample from 2 clinics in Kampala, Uganda. To meet inclusion criteria, patients had to be HIV-infected adults experiencing a pruritic skin eruption for greater than 1-month duration, with evidence of multiple papular or nodular lesions. The study was performed over a 3-week period from May 19 through June 6, 2003. Analysis of biopsy specimens occurred in the United States after the 3-week study period.

In the month leading up to the study, participants meeting inclusion criteria and willing to participate were referred by physicians and nurses from the 2 Ugandan clinics (each clinic was asked to recruit consecutive qualifying patients until 75 were referred). One clinic was an HIV clinic in a large public hospital (Mulago Hospital), and the other was a community-based, nongovernmental organization providing care to nearby residents (Reach-Out Mbuya Parish HIV/AIDS Community Initiative). The patient volume at each clinic was about 300 visits per week. Together, these 2 clinics care for a broad spectrum of patients living with HIV in urban Uganda. At the time of the study, both clinics were offering treatment for opportunistic infections, but neither was providing antiretroviral therapy due to lack of resources. Referring physicians indicated that for the Mulago hospital clinic, where staff kept count, only 2 patients who met inclusion criteria during the enrollment period declined to participate. Staff at the Reach-Out clinic did not keep count but reported that fewer than 5 patients meeting inclusion criteria did not participate. The study provided the opportunity for patients to access additional care at no cost, and the 2 sites did not experience significant difficulty in meeting recruitment goals.

This study was carried out with approval from institutional review boards at the University of California, San Francisco, and Makerere University in Uganda. Referred patients received information in their own language (including English, Acholi, and Luganda) explaining that the study was an effort to determine the cause of an itchy skin rash associated with HIV. Trained counselors reviewed written consent forms that detailed all aspects of the study, as well as the potential risks and benefits, with every referred patient in his/her preferred language. Every patient willing to participate provided written informed consent before undergoing any aspect of the study. The consent forms, approved by both institutional review boards, indicated that no individual identities would be used in published reports resulting from the study. The consent form included sections on the purpose, background, procedures, risks, benefits, and alternatives. The procedures section included paragraphs on the questionnaire, blood tests, medical record reviews, skin examinations, skin biopsies, skin cultures, suture removals, and receipt of test results. A section on confidentiality stated that “Participation in research will mean some loss of privacy. However, my research records will be handled as confidentially as possible. All research records will be coded so that the laboratories and researchers will not know which samples are mine. My clinic doctor will have access to the codes so that I can be told the results of my specific diagnostic tests. No individual identities will be used in any reports or publications that result from this study.” An additional section on the voluntary nature of the study stated that “If I agree to participate in this study, I may stop participating at any time. I have the right to participate or withdraw at any point in this study without the risk of losing my medical care.” Those patients whose photographs were taken signed a separate consent form giving permission for reproduction of images in published reports.

After the extensive consent process, enrollees were administered an epidemiologic questionnaire by clinic nursing staff in 1 of 3 languages (English, Acholi, and Luganda) preferred by the patient. The staff received training in question administration prior to the study. The questionnaire included items on patient demographics, history of the rash and attempted treatments, environmental exposures, and medication history including traditional therapies. All patients had medical charts available for review, and chart review was used to ascertain past medical conditions including HIV-associated opportunistic infections and tuberculosis, as well as to supplement information about medication use. Patients were also asked about previous infections and medication use during the physical examination, to supplement the chart review. A 12-month cutoff was applied when eliciting use of medications.

Each participant received physical examinations by 2 US-based dermatologists (J.S.R., M.V.) that included consensus subjective assessments of rash severity. The rash of each participant was rated by both examining physicians (who were blinded to the results of any laboratory or biopsy results) until verbal consensus was reached. No accepted objective severity scale for PPE is in use, and the authors did not develop one for use as a part of this study. Mild eruptions consisted of sparse lesions (<20), usually limited to the distal extremities, with only minimal scarring or pigmentary changes. Moderate eruptions were characterized by approximately 20 to 60 lesions over the trunk and extremities, usually with some degree of excoriation, pigmentary alteration, and scarring. Patients with severe eruptions had high lesion counts (usually >60) with marked excoriation, pigmentary alteration, and scarring and, occasionally, facial involvement.

Participants also had laboratory studies, including serologic testing to confirm HIV infection, complete blood cell count with differential, and CD4 cell count. The serologic tests for HIV were performed using parallel testing with 2 kits (Thermo Labsystems HIV [Ani Labsystems Ltd, Vantaa, Finland] and Abbott-Murex HIV [Abbott Laboratories, Abbott Park, Ill]) with tie-breaking of discordant results using Western blot (though no tie-breaks were required). CD4 cell counts were performed using FACSCount (Becton Dickinson, La Jolla, Calif). Although serology results and CD4 cell counts were obtained on all participants, specimen amounts were inadequate to report automated differential blood counts in 16 patients. Viral loads were not obtained because of limited resources and because, in the setting of the limited budget, we determined that the CD4 cell counts were a more reliable marker of immunological status and integrity, given that only a single laboratory measurement was being obtained at a single point.

After local anesthesia, a 4-mm punch biopsy of a recent primary skin lesion was performed. During the skin examination, patients were asked to identify lesions that were new and that, based on their experience, would develop into more advanced PPE lesions. Among these lesions, the investigators selected 1 indurated papule without excoriation for biopsy.

Tissue samples were formalin-fixed and shipped to the United States for analysis by a senior dermatopathologist (P.E.L.) with recognized expertise in both dermatopathology and HIV pathology; this investigator was blinded to any patient status that could have biased interpretation of the slides. Six sections from each specimen were mounted on a slide and stained with hematoxylin-eosin for primary light microscopic analysis. Additional level sections were obtained if inflammatory infiltrates were not present in the initial sections or if infiltrates were present around pilosebaceous units but the pilar canal of the affected follicle was not adequately displayed. Only 1 specimen required additional level sections. Special stains for organisms (including periodic acid–Schiff [digested], Brown-Brenn, Fite, and Warthin-Starry) were available for the study. These were not performed on any cases, however, as none had histopathologic findings of an infection (eg, suppurative, suppurative and granulomatous, or granulomatous dermatitis).

Statistical analyses were performed using SPSS version 11.5 (SPSS Inc, Chicago, Ill) and Stata version 8.0 (StataCorp, College Station, Tex). We used t tests and linear regression (assessing a linear contrast of regression coefficients for trend) to evaluate differences in age; χ² tests (including tests of trend) to evaluate differences in proportions such as sex, clinic, severity, and biopsy findings; and nonparametric tests (including the Kruskal-Wallis test and the Cuzick extension to the Wilcoxon rank-sum test for trend) to evaluate differences in CD4 cell and eosinophil counts, which were abnormally distributed. We used logistic regression to adjust for differences in the CD4 cell count when comparing the likelihood of having certain biopsy findings between clinic sites. χ² Tests were used when comparing the frequency of absolute eosinophilia or the frequency of certain biopsy findings among subgroups of patients. The only missing data were eosinophil counts in the 16% (16/102) of patients with inadequate blood samples, and those patients were excluded only from analyses involving peripheral eosinophil counts. Also, 1 patient was missing data on rash distribution. P<.05 was used to determine statistical significance.

One hundred two (77.3%) of 132 patients referred completed the study from May 19 through June 6, 2003. The other 30 were excluded because they did not meet inclusion criteria when evaluated by the study dermatologists (23 [17.4%]) or left the clinic before completing the study (7 [5.3%]). The 102 patients who finished the study represented 93.6% of the 109 meeting inclusion criteria. The mean age of patients completing the study was 35 (SD, 8) years, and the majority (83 [81%]) were women. Results of serologic testing to confirm HIV infection were positive in all patients, and CD4 cell counts in this population were low (median, 46/μL; interquartile range, 9-146/μL).

Fifty-two patients were from the hospital-based clinic, and 50 were from the community-based clinic. Demographic characteristics of the 2 groups were similar, including age and sex (Table 1). The patients from Mulago Hospital, however, had significantly lower CD4 cell counts (median, 33 vs 78/μL; P = .04).

On subjective global assessment by the investigators, patients were described as having mild (30 [29%]), moderate (53 [52%]), or severe (19 [19%]) rashes. Mild eruptions consisted of fewer than 20 lesions, most commonly in an acral distribution. Typical patients with moderate eruptions had approximately 20 to 60 excoriated, hyperpigmented papules and nodules on the trunk and extremities. Those with severe eruptions had high lesion densities over larger body surface areas, often including the face, trunk, and all 4 extremities. New lesions were typically characterized by a 2- to 8-mm erythematous papule. Subacute lesions were often excoriated papules, while chronic lesions were larger papules to nodules (approximately 1 cm) with marked lichenification and hyperpigmentation (Figure 1). Increasing rash severity was significantly associated with lower CD4 cell counts (P<.001) and higher absolute peripheral eosinophil counts (P = .01) (Table 2).

The majority of patients (n = 86 [84.3%; 95% confidence interval, 77%-91%]) had pathologic skin biopsy findings characteristic of arthropod bites. Most of these (n = 53) had roughly wedge-shaped, moderately dense to dense, superficial and deep, perivascular and interstitial infiltrates of lymphocytes and many eosinophils beneath an epidermis that was slightly hyperplastic (Figure 2). The infiltrates often extended into the subcutis. Varying numbers of neutrophils were also present. These are the features classically seen in arthropod bites or stings.¹⁴ Some (n = 11) also showed a punctum (a small focus of spongiosis positioned in the epidermis over the center of the dermal infiltrate) at the site of the bite, which is nearly pathognomonic. Biopsy findings from the remaining 33 of the 86 patients were described as having dermal perivascular and interstitial infiltrates of lymphocytes and eosinophils—findings strongly suggestive of arthropod bites.

Most of the remaining biopsies (n = 10 [9.8%]) showed nonspecific inflammatory changes (perivascular lymphocytes). The other 6 (5.9%) included 1 case each of a prurigo nodule, lichen simplex chronicus, scar, postinflammatory pigmentary alteration, spongiotic dermatitis, and suppurative folliculitis. There was not a single case of eosinophilic folliculitis.

There were differences between the groups of patients with and without biopsy findings suggestive of arthropod bites. Compared with patients without histological evidence of arthropod bites on biopsy, those with bites had significantly higher peripheral eosinophil counts (median, 180 vs 330/μL, respectively; P = .02), and had a trend toward lower CD4 cell counts (median, 99 vs 40/μL; P = .07). Malaria and tuberculosis are prevalent in the area and some patients had been treated for these conditions, but we did not believe that their prevalence would have an effect on differences in eosinophilia among study groups. Also, although we had the ability to obtain scrapings for dermatophytes and scabies, none of the study patients had clinical findings supporting either of these and so no scrapings were performed.

The physical distribution of the rash was not predictive of biopsy results showing arthropod bites (arthropod bite on biopsy in 81% [13/16] of patients with bites on extremities only, 85% [44/52] of those with bites on extremities and trunk, and 88% [29/33] of those with bites on extremities, trunk, and face; P = .82; missing distribution data for 1 patient). The frequency of arthropod bites on biopsy was the same among men (84% [16/19]) and women (84% [70/83]) in the study group (P = .99). Forty-four percent of patients had active or previously treated pulmonary tuberculosis (TB), but TB status was not predictive of biopsy results (arthropod bite on biopsy in 82% [47/57] of patients without known TB, 88% [21 of 24] of those with successfully treated prior TB, and 86% [18/21] of those with active TB; P = .83). Patients from the public hospital clinic were more likely to have arthropod bites on biopsy than were patients from the community-based clinic (92% vs 76%, P = .02), and there was a trend in this regard even when controlling for CD4 cell count (P = .06).

The systemic medications most frequently prescribed within the previous 12 months, in descending order, were paracetamol/acetaminophen (59 patients), trimethoprim/sulfamethoxazole (51), promethazine (38), ketoconazole (34), amitriptyline (23), doxycycline (23), and pyrimethamine/sulfadoxine (16). None of the patients was receiving antiretroviral therapy.

The topical medications most frequently prescribed within the prior 12 months included calamine (41 patients), clotrimazole (33), topical steroids (21), salicylic acid (20), and Whitfield ointment (14). Systemic medication history was not associated with clinical rash severity or biopsy results in the study participants (P>.05 for all, by χ² test). For topical steroids, however, users were more likely to have an arthropod bite on biopsy (21/21 vs 65/81 [80%] for users vs nonusers, respectively; P = .03). We believe this may be due to chance, because topical steroids are widely used on biopsied lesions and their effects on histology are widely appreciated; if anything, they could mask the immune cell response in the sections but would not likely increase the likelihood of reactions to arthropod bites on biopsy. Also, most nonusers of topical steroids had biopsy findings of arthropod bites. The use of Whitfield ointment was also associated with biopsy findings. Nine (64%) of the 14 users had an arthropod bite on biopsy, as did 77 (88%) of the 88 nonusers (P = .03). Pathologist investigators could not hypothesize a way in which the ointment (benzoic acid and salicylic acid) could have affected the likelihood of finding reactions to arthropod bites on biopsy and believed the small numbers of users weighed against a statistically interesting finding. Furthermore, it would be difficult to argue that either of the topical treatments caused the biopsy findings of arthropod bites, because most nonusers of each ointment had biopsy findings consistent with arthropod bite.

Per questionnaire responses, 76% of study patients reported exposure “to many bugs or insects at home/work,” and 56% reported sustaining many bites. Of those reporting frequent bites, the majority (84%) listed mosquitoes as the offending agent.

Our findings suggest that the PPE occurring in HIV-infected individuals in Uganda may be a reaction to arthropod bites. Most of the biopsy specimens of early lesions of PPE showed findings characteristic of arthropod bites—namely, moderately dense to dense, superficial and deep, perivascular and interstitial infiltrates of lymphocytes and eosinophils often extending into the subcutis, with epidermal hyperplasia and, in some cases, accompanied by a punctum. We thus suggest that “arthropod-induced prurigo of HIV” may be a more specific and appropriate description of this clinical entity than PPE.

The lesions we identified as arthropod bites are not consistent with any other diagnoses. The infiltrates were not centered around hair follicles, arrector pili muscles, or sebaceous lobules (as in eosinophilic folliculitis). Although drug eruptions can mimic almost every pattern of inflammatory skin disease and have overlapping features with reactions to arthropod bites, we do not believe that PPE is a drug reaction. The infiltrates in most drug eruptions are not as dense as those seen in our cases and do not extend into the subcutis. Puncta would not be seen in drug eruptions. The findings were not compatible with conditions that can simulate PPE clinically, such as staphylococcal folliculitis (suppurative folliculitis, demonstrated histologically), Demodex folliculitis (spongiotic infundibular folliculitis), phototoxic or photoallergic dermatitis (many single necrotic keratinocytes or spongiotic dermatitis, respectively), crusted scabies (mites, fecal deposits, or eggs beneath the stratum corneum), secondary syphilis (a superficial and deep, psoriasiform lichenoid pattern with lymphocytes, plasma cells, and histiocytes predominating), onchodermatitis (a deep zone of fibrosis with microfilariae), or papulonecrotic tuberculid (a necrotizing granulomatous reaction).

By aggressively seeking new primary lesions instead of older, scarred, or excoriated nodules, we were able to obtain diagnostic biopsy results in a majority of patients. The patients were adamant that the lesions biopsied were representative of the initial presentation of all their PPE skin lesions. We thus strongly doubt that the biopsied lesions represent a concurrent process that was unrelated to the patients’ more advanced PPE lesions.

There are a number of possible explanations for the different findings observed in the biopsies of the 16 patients without histopathological evidence of arthropod bites. In 4 of these patients, the results suggested that an older or secondary lesion was inadvertently selected for sampling. These 4 biopsies may represent prior arthropod bite reactions in later stages. Many of the rest had highly nonspecific inflammatory changes. It is possible that these lesions were also arthropod bites, as it is not uncommon to have these nondiagnostic findings in known cases of arthropod-induced injuries.¹⁵

On the other hand, the patients without diagnostic biopsy findings might have had some cause other than arthropod bites for their skin lesions. The nonbite subgroup had a trend toward higher CD4 cell counts and had significantly lower eosinophil counts. While these differences could suggest a different etiology, they might also indicate an earlier stage of HIV in which lesions diagnostic for arthropod bites are more difficult to detect. The different laboratory profiles might also simply reflect a variation in the immunological reaction pattern in response to the same causative insult.

The diffuse distribution of PPE lesions (most dense on the extremities, but also in areas that are typically covered) raises the issue of whether bites are the initiating event for every lesion. In our study patients, biopsies from the trunk were no less likely to show primary arthropod bites than were those from the extremities. In addition, many arthropods (including mosquitoes) are known to bite through clothing and on the trunk. We believe that most (if not all) PPE lesions do begin as arthropod bites. We did not, as part of this study, attempt to identify the species responsible for the patients’ bites or stings. To do so would be difficult because patients usually do not feel discomfort during the blood meal of certain arthropods (including mosquitoes)^16- 18 and because the histopathology of bites and stings from different arthropods is generally indistinguishable. Nevertheless, most study patients reporting frequent arthropod bites did list mosquitoes as the source of their bites.

The association between arthropod bites and PPE raises many pathophysiological issues. Patients with PPE may lose the well-described desensitization^19- 20 most people develop after repeated exposure to arthropod bites. The experience of increasing hypersensitivity to previously tolerated medications in HIV-infected patients as CD4 cell counts decline²¹ provides one possible model. Some have hypothesized that this profound increase in drug hypersensitivity²² could be secondary to polyclonal B-cell activation²³ or T-cell activation²⁴ seen with AIDS. These same hypothesized phenomena could be involved as patients with PPE become hypersensitive to arthropod antigens they may have previously tolerated.

The idea that arthropod bite hypersensitivity might account for the pruritus observed in some HIV-infected patients has been raised previously, though the literature on this subject is sparse. One study of 7 AIDS patients in Florida with pruritus (though not classic PPE) found that 5 had increased antibody titers to antigens in the saliva of a common local mosquito.²⁵ The authors hypothesized that the pruritus in some patients with AIDS might be a form of chronic “recall” reaction to mosquito salivary antigens in the setting of nonspecific B-cell activation. Another study found that patients with PPE in Brazil had lower IL-2 and γ-interferon levels than did HIV-infected patients without dermatitis,¹⁰ raising the possibility of a potential role of T_H2-dominant humoral immunity. A shift from T_H1 to T_H2 cytokine predominance has been associated with advancing HIV.^26- 28 A subset of patients with pronounced T_H2 dominance may be more predisposed to the development of the arthropod hypersensitivity leading to PPE.

The HIV-infected patients in industrialized nations who have intractable pruritus²⁹ and who have prurigo nodularis³⁰ frequently have accompanying hypereosinophilia. Prurigo nodularis, which can clinically resemble mild PPE, is an idiopathic condition in which nodular lesions result from repetitive picking or scratching. Though it may occur in some HIV-infected patients, it may also be associated with renal failure and hepatic dysfunction and may be seen in some patients without other medical problems. In our study population, higher eosinophil counts were associated with more severe clinical PPE eruptions. Since the IL-4 and IL-5 produced by T_H2 lymphocytes promote IgE production and eosinophil differentiation,³¹ this further supports the idea that a subset of patients with advanced HIV infection and hyperactive T_H2 immune response may be more likely to experience exaggerated immunological phenomena in the skin.

Another question remains unanswered, ie, why is this condition more prevalent in Africa (and possibly southeast Asia) than in North America and Europe? One possibility is that arthropod antigens on those continents are simply different—ie, more likely to cause this reaction pattern. Another possibility is that host factors predispose certain populations. One study of HIV-infected patients with rashes in the southern United States found that eosinophilia and skin rashes were more common in African Americans.³⁰

Some researchers are suggesting that PPE be added to the list of conditions qualifying patients for initiation of HIV medications in settings that rely on clinical indicators of immune suppression.³² Many of our patients were enrolled in trials of antiretroviral therapy after completing the skin study, and anecdotal follow-up suggests that their symptoms of PPE are improving. More formal work is under way to establish the course of this eruption after initiation of antiretroviral agents. Determining the effectiveness of HIV therapy for the symptoms of PPE might also provide additional information on the immunological basis of this condition.

Follow-up studies may be required to determine whether these results are generalizable to other places where this eruption has been reported (eg, other African countries, southeast Asia, and the Caribbean). This study, however, provides evidence to support the hypothesis that PPE is actually an arthropod-induced prurigo representing an exaggerated immune response to arthropod antigens in predisposed HIV-infected patients.