Nail Dystrophy, Edema, and Eosinophilia: Harbingers of Severe Chronic GVHD of the Skin in Children

ORIGINAL ARTICLE Nail dystrophy, edema, and eosinophilia: harbingers of severe chronic GVHD of the skin in children

JT Huang1, CN Duncan2, D Boyer3, H Khosravi4, LE Lehmann2 and A Saavedra5

The prognostic value of adnexal findings in chronic GVHD (cGVHD) has not been investigated in children. Dermatologic examinations were performed in a severe cohort of 11 children with skin cGVHD seen over a 2-year period. Findings were compared with 25 additional patients with skin cGVHD and 97 control patients. In 36 patients with skin cGVHD, nail dystrophy was present in 45% of patients, and was significantly associated with sclerotic disease and lung cGVHD. Pterygium inversum unguis (PIU) was associated with severe lung disease, with significantly lower % predicted FVC and FEV1 in those with PIU than those without. Forty-four percent of GVHD patients had preceding peripheral edema and 56% had preceding peripheral eosinophilia. Peripheral edema and eosinophilia were significantly associated with sclerotic cGVHD and persisted until the diagnosis of cGVHD in all patients. Comparison of data with control patients showed that incidence of nail dystrophy, incidence of peripheral edema and mean peak peripheral eosinophil count of patients with skin cGVHD was significantly higher than those without cGVHD. This study suggests that nail dystrophy, persistent peripheral edema and persistent peripheral eosinophilia are harbingers of severe cGVHD of the skin in children. The presence of PIU may be a harbinger of severe lung involvement.

Bone Marrow Transplantation (2014) 49, 1521–1527; doi:10.1038/bmt.2014.194; published online 22 September 2014

INTRODUCTION seen by pediatric dermatologists at Dana-Farber Cancer Institute/Boston ’ Chronic GVHD (cGVHD) is a significant cause of morbidity and Children s Hospital over a 2-year period (February 2011 through March – 2013). Retrospective chart review of pediatric HSCT recipients transplanted mortality in children, second only to relapse of disease.1 3 4 between 2006 and 2012 was also performed. IRB approval through the Mortality from pediatric cGVHD has been estimated at 41%. Office of Human Research Studies at DFCI was obtained. However, prognostic indicators for morbidity and mortality For the cross-sectional cohort study, inclusion criteria included age at secondary to cGVHD have not been clearly identified in children. HSCT ⩽ 18 years, recipient of allogeneic HSCT, and the diagnosis of skin The skin is the most commonly affected organ in cGVHD and its cGVHD. Exclusion criteria included recipient of ⩾ 2 HSCT, and HSCT involvement may be predictive of a poor prognosis.5 Chronic skin performed at an outside institution. For the diagnosis of cGVHD, patients GVHD may present in multiple different ways ranging from xerosis either met diagnostic criteria established by the 2005 NIH Consensus Development Project, or had a new onset persistent skin eruption to erythroderma to sclerotic changes. Adnexal structures may also 8 be involved in cGVHD, resulting in nail dystrophy or alopecia. consistent with a feature listed in this report. Informed consent was Clinically, we have observed that patients with adnexal obtained by the patient and/or legal guardian. During their dermatology visit, eligible patients were recruited and involvement are less likely to respond to standard therapies. The enrolled in the study, which included a one time evaluation of their skin. nail matrix and hair follicle are known to be sites of immune Participants had a complete examination of the skin, mucous membranes, privilege, with decreased HLA-A, B and C expression on hair and nails by at least one of the two dermatologists (authors JTH, AS). keratinocytes and melanocytes as well as decreased MHC class II All abnormalities were documented and photographed. Patients and their – expression of epidermal antigen presenting cells.5 7 Downregula- legal guardians completed age-appropriate PedsQL questionnaires.9 tion of antigen recognition and T-cell activation decreases the Medical records of participants were reviewed and additional data were likelihood of cGVHD at these sites. Thus adnexal involvement may gathered, including general demographic data (age, gender), primary indicate more profound immune dysregulation and predict disease/reason for HSCT, source of HSCT, conditioning regimen, HLA treatment-refractory disease. match, gender of donor, history of acute GVHD (aGVHD), past and current The primary goal of our study was to investigate whether the immunosuppressive medications. Laboratory results and relevant radio- graphic studies were reviewed. When available, lowest pulmonary function presence of adnexal involvement in skin cGVHD was associated test (PFT) data (%predicted FVC and FEV1) were recorded. with severe disease. We also examined the incidence of edema Medical records of all pediatric patients who received an unrelated HSCT and eosinophilia as potential early signs of skin cGVHD. Finally, between May 2006 and August 2012 (date range HSCT was performed in we provide a descriptive analysis of a severe cohort of pediatric study cohort) were also reviewed. We gathered data on two separate patients with skin cGVHD. populations: (1) those with a history of skin cGVHD and (2) those without any history of cGVHD (control group). Patients excluded from this analysis MATERIALS AND METHODS included those with ⩾ 2 HSCT (including autologous HSCT), double cord transplants, death due to acute complications, and relapse of primary Study design condition. In patients with skin cGVHD, any documentation of nail This study was a single-center cross-sectional cohort study of pediatric dystrophy, eosinophil counts, and peripheral edema around time of hematopoietic SCT (HSCT) recipients with a diagnosis of skin cGVHD cGVHD diagnosis was recorded. In control patients, any documentation of

1Dermatology Program, Boston Children’s Hospital, Boston, MA, USA; 2Department of Pediatric Oncology, Dana-Farber Cancer Institute, Boston, MA, USA; 3Division of Respiratory Diseases, Boston Children’s Hospital, Boston, MA, USA; 4 Harvard Medical School, Boston, MA, USA and 5Department of Dermatology, Brigham and Women’s Hospital, Boston, MA, USA. Correspondence: Dr JT Huang, Dermatology Program, Boston Children’s Hospital, Fegan 6, 300 Longwood Avenue, Boston, MA 02115, USA. E-mail: [email protected] Received 3 March 2014; revised 3 July 2014; accepted 17 July 2014; published online 22 September 2014 Harbingers of severe skin cGVHD in children JT Huang et al 1522

Table 1. Characteristics of skin cGVHD in patients transplanted between 2006–2012

Characteristics Study cohort (11) Nonstudy cohort (25) P-valuea Total (36)

History of acute cutaneous GVHD 1.0 Yes 5 (45%) 10 (40%) 15 (42%) No 6 (55%) 15 (60%) 21 (58%) Time from HSCT to cGVHD 0.94 Mean in days (s.d.) 245 (92) 249 (172) 248 (151) Onset of cGVHD 0.70 Progressive 4 (36%) 6 (24%) 10 (28%) Quiescent 1 (9%) 5 (20%) 6 (17%) De novo 6 (55%) 14 (56%) 20 (56%) Onset of skin involvement 0.7 Presenting organ 7 (64%) 18 (72%) 25 (69%) Subsequent organ 4 (36%) 7 (28%) 11 (31%) Type of skin cGVHD 0.025 Sclerotic 8 (73%) 7 (28%) 15 (42%) Nonsclerotic 3 (27%) 18 (72%) 21 (58%) Steroid responsiveness 0.15 Steroid responsive 3 (27%) 15 (60%) 18 (50%) Steroid resistant 8 (73%) 10 (40%) 18 (50%) Extracutaneous involvement 0.69 None 3 (27%) 8 (32%) 11 (31%) Mucous membranes 1 (9%) 4 (16%) 5 (14%) Visceral organs 1 (9%) 5 (20%) 6 (17%) Mucous membranes and visceral organs 6 (55%) 8 (32%) 14 (39%) Lung cGVHD 0.22 Yes 5 (45%) 5 (20%) 10 (28%) No 6 (55%) 20 (80%) 26 (72%) Death at time of publicationb 1.0 Yes 4 (36%) 9 (36%) 13 (36%) No 7 (64%) 16 (64%) 23 (64%) aP-values based on Fisher’s exact test for categorical variables and two-sample t-test for continuous variables. bJune 2014.

nail dystrophy, eosinophil counts and peripheral edema from 100 days RESULTS post-HSCT to May 2013 was recorded. In controls, high eosinophil counts Demographics attributed to acute GVHD were excluded. Eleven pediatric HSCT recipients with skin cGVHD were enrolled in the cross-sectional cohort study (study cohort). These patients Study aims were transplanted between May 2006 and August 2012. There The primary aim of our study was to investigate whether adnexal were an additional 25 patients with skin cGVHD who were involvement, including nail dystrophy and alopecia, in skin cGVHD was transplanted during this time period, 12 of whom had active skin predictive of more severe disease. Secondary aims of our study were to disease during the study time period of 2011 to 2013. Of these identify early signs of skin cGVHD, and to provide a descriptive analysis of 12 patients, 10 were noted to have eczematous or erythematous our cohort. skin eruptions, and were not seen by dermatologists during the study period. The remaining two had sclerotic disease; one declined study enrollment, and the other was critically ill and Study definitions fi deemed unsuitable for the study. Date of cGVHD was de ned as onset of symptoms that were ultimately Between May 2006 and August 2012, there were 209 recipients diagnostic of cGVHD. Determination of overall cGVHD severity was based of unrelated HSCT performed at our institution. Of these 209 on the scoring system proposed by 2005 NIH Consensus Development 8 patients, 18 had received ⩾ 2 HSCT (including autologous HSCT), Project. Fully matched patients had no identified mismatches at HLA-A, -B, -C and –DRB1 by high resolution typing if the graft source was BM or no 1 had a double cord transplant, 25 died of acute complications, 21 identified mismatches at HLA-A, -B, -DRB1 by intermediate resolution had subsequent relapse of their primary condition and 47 patients typing if the graft source was cord blood. De novo onset type of cGVHD had chronic GVHD (36 of whom had skin involvement). The was defined as no prior history of acGVHD, quiescent onset type of cGVHD remaining 97 patients were included in our control group. was defined as history of aGVHD with absence of immunosuppressive Demographic data of patients with skin cGVHD and controls are medications at the time of onset of cGVHD, and progressive onset type summarized in Supplementary Appendix A. Control patients were of cGVHD was defined as history of aGVHD with presence of significantly younger than cGVHD patients. Otherwise, there were immunosuppressive medications at the time of onset of cGVHD.10,11 no significant differences in patient demographics.

Data analysis Characteristics of cGVHD Characteristics of each group were summarized using descriptive methods. Characteristics of cGVHD in all 36 patients with skin cGVHD are Comparisons of categorical variables, such as peripheral edema, nail summarized in Table 1. Overall, 42% (15/36) of patients had dystrophy and PIU, were based on Fisher’s exact tests; comparisons of sclerotic cGVHD, 69% of patients had skin manifestations upon continuous variables, such as peripheral eosinophil counts, were based on initial presentation of cGVHD, 69% of patients had extracutaneous two-sample t-tests. Statistical signiﬁcance was deﬁned as Po0.05. There involvement, including 39% with involvement of both mucous were no adjustments for multiple comparisons. membranes and visceral organs. Patients with sclerotic cGVHD

Table 2. Characteristics of sclerotic versus nonsclerotic cGVHD

Characteristics Sclerotic disease (15) Nonsclerotic disease (21) P-valuea

Steroid resistant disease 0.0001 Yes 87% (13/15) 19% (4/21) No 13% (2/15) 81% (17/21) Lung cGVHD 0.06 Yes 47% (7/15) 14% (3/21) No 53% (8/15) 86% (18/21) Death at time of publication 0.02 Yes 60% (9/15) 19% (4/21) No 40% (6/15) 81% (17/21) Nail dystrophy 0.0005 Yes 73% (11/15) 14% (3/21) No 20% (3/15) 67% (14/21) Unknown/missing 7% (1/15) 19% (4/21) Peripheral edema o0.0001 Yes 87% (13/15) 14% (3/21) No 13% (2/15) 86% (18/21) Mean peripheral eosinophil count in cells/μL (s.d.) 1570 (1040) (N = 14) 840 (1170) (N = 20) 0.07 Eosinophiliab 0.008 Yes 80% (12/15) 33% (7/21) No 13% (2/15) 62% (13/21) Unknown 7% (1/15) 5% (1/21) aP-values based on non-missing categories. Comparison of sclerotic versus nonsclerotic disease for nail dystrophy and peripheral edema based on Fisher’s exact test; P-value for eosinophil count based on two-sample t-test. bEosinophilia defined as peripheral eosinophil count 4500 cells/μL. were significantly more likely than those with nonsclerotic skin patients with nail dystrophy (9/14) had findings of longitudinal involvement to have steroid-resistant disease (13/15 versus 4/21, ridging and distal splitting of the fingernails. Specific nail findings P = 0.0001). At the time of publication, a significantly higher are listed in Supplementary Appendix B. Analysis of nail dystrophy proportion of patients with sclerotic disease than those with was limited to study and nonstudy patients with documented nonsclerotic skin disease had subsequently died of treatment nail exams. related causes (9/15 versus 4/21, P = 0.02; Table 2). Patients with nail dystrophy were significantly more likely than Patients in the study cohort were significantly more likely than those with normal nails to have sclerotic disease (11/14 versus nonstudy cGVHD patients to have sclerotic cGVHD (8/11, 73% 3/17, P = 0.001). Patients with nail dystrophy were also more likely versus 7/25, 28%, P = 0.025). There were no significant differences to have steroid-resistant disease and lung cGVHD (11/14 versus in extracutaneous organ involvement, steroid responsiveness or 3/17, P = 0.001, and 8/14 versus 2/17, P = 0.02, respectively). death rates between cohorts (Table 1). Twenty-three of the 31 patients with documented nail exams had PFTs performed at our institution. Analysis of PFT results Dermatologic assessment of study cohort revealed that patients with nail dystrophy had significantly lower Median time from initial development of skin cGVHD to time of % predicted FVC and FEV1 than those without nail dystrophy study enrollment was 196 days (range 48–1617 days). Eight of 11 (Table 3). patients in the study cohort had sclerotic cGVHD. Of these eight Within the study cohort, patients with PIU were significantly patients, five had lichen sclerosus, dermal fibrosis and myofascial more likely than those without PIU to have lung cGVHD involvement; two had myofascial involvement alone; and one had (5/5 versus 0/6, P = 0.002, Table 3). All five patients with PIU had dermal fibrosis and myofascial involvement. Lichen sclerosus-like evidence of significant lung cGVHD. Analysis of PFT results changes occurred predominantly at flexural regions. One patient revealed that patients with PIU had significantly lower % predicted had penile lichen sclerosus et atrophicus (Figure 1). The three FVC than those without PIU (20.4 versus 85%, Po0.0001). patients without sclerotic features had eczematous/follicular Similarly, patients with PIU had significantly lower % predicted cGVHD (two) and ichthyosiform cGVHD (one; Figure 2). Dyspig- FEV1 than those without PIU (14.6 versus 85%, Po0.0001). Of the mentation (both hyperpigmentation and hypopigmentation) was five patients with PIU, four were referred for lung transplant a prominent feature in all the 11 patients. Vitiliginous changes, evaluation and the remaining patient was screened and not including skin depigmentation and poliosis, were noted in four considered a candidate due to her severe multisystem disease. patients with sclerotic cGVHD. Lung biopsy results of four patients with PIU were reviewed, Seven patients had some form of nail dystrophy noted at the showing varying degrees of cGVHD. PIU was also significantly study visit. Five patients had nail findings consistent with associated with a lower quality of life based on self report using pterygium inversum unguis (PIU, Figure 3). The two remaining the PedsQL questionnaire (PedsQL score 51.4 versus 69.3, patients, had longitudinal ridging and splitting of their nails. In all P = 0.009). Due to lack of detailed documentation of nail findings fi the patients, nail changes predominantly involved ngernails. in nonstudy patients, analysis of PIU was limited to study patients. Eight patients had scalp involvement with scaling of the scalp There was sufficient documentation of nail findings in only five and/or hair thinning. patients to inform a temporal relationship between nail dystrophy and cGVHD. Three of these patients had nail dystrophy occurring Adnexal involvement in skin cGVHD 0, 3 and 5 months before sclerotic disease and two patients had There were 31 patients with documented nail exams (11/11 in nail dystrophy occurring 2 months and 5 years before the onset of study cohort and 20/25 in nonstudy cohort). Of these 31 patients, severe lung cGVHD. Of note, four of five study patients with PIU there were 14 patients with nail dystrophy. The majority of had no prior documented nail exam.

Figure 1. Spectrum of morphology in sclerotic cGVHD. (a) Lichen sclerosus-like disease with profound dyspigmentation. (b) Myofascial disease. (c) Penile lichen sclerosus. (d) Dermal sclerotic plaques in ‘choke-hold’ distribution. (e, f) Bound-down sclerosis of toes and guttate sclerotic papules and plaques.

Figure 2. Two patients with new onset, persistent and generalized dermatitis after HSCT, consistent with eczematous/ichthyosiform cGVHD. (a) Note unusual folliculocentric nature and associated dyspigmentation. (b, c) Sudden and generalized ichthyosis 154 days after HSCT.

Medical record review identified three patients transplanted toenails. The incidence of nail dystrophy in cGVHD patients was between 2006 and 2012 who developed lung cGVHD without significantly higher than that of the control group (Table 4). skin involvement. Notably, all the three patients had normal nail exams. Peripheral edema in skin cGVHD A total of 16 patients (6 in the study cohort; 10 in the nonstudy Control group. Of the 97 control patients, no nail exam was cohort) were noted to have otherwise unexplained edema of the documented in 29, normal nails were documented in 63 and lower ± upper limbs before the onset of skin cGVHD. One study dystrophic nails were documented in 5 patients. Of the five patient also had scrotal edema (Figure 4). Thirteen of 16 patients patients with nail dystrophy, two had a primary diagnosis of subsequently developed sclerotic cGVHD with mean time from dyskeratosis congenita, and one had a primary diagnosis of onset of edema to evident sclerotic disease of 58.5 days (range Fanconi anemia, all with nail abnormalities before HSCT. Of the 10–243). The patient with scrotal edema developed lichen remaining two patients, one had documentation of ‘distal nail sclerosus of the glans penis. All 13 patients were noted to have dystrophy’ at a single visit that had resolved by the following visit, sclerotic changes immediately upon resolution of edema. Of note, and one had documentation of thickened and discolored great the remaining two patients with sclerotic disease who did not

Figure 3. Pterygium inversum unguis in four patients with severe cGVHD characterized by sclerotic involvement and lung disease.

Table 3. Nail dystrophy in cGVHD

Nail ﬁnding (n) Steroid resistant disease Sclerotic cGVHD Lung cGVHD Mean % predicted FVC Mean % predicted FEV1 (14,8) (14,8) (10,5) (23,10)a (23,10)a

Nail dystrophy (14)b 11/14 11/14 8/14 46 (s.d. 29) 42 (s.d. 34) No nail dystrophy (17) 3/17 3/17 2/17 91 (s.d. 16) 90 (s.d. 17) Fisher’s exact P-value 0.001 0.001 0.02 0.0002 0.0004 PIU (5)c 5/5 5/5 5/5 (100%) 20.4 (s.d. 9.1) 14.6 (s.d. 7.9) No PIU (6) 3/6 3/6 0/6 (0%) 85.0 (s.d. 16.5) 85.0 (s.d. 14.5) Fisher’s exact P-value 0.18 0.18 0.002 Two-sample t-test P-value o0.0001 o0.0001 Abbreviations: PFT = pulmonary function test; PIU = pterygium inversum unguis. aNo PFT data on eight patients in entire skin cGVHD cohort, one patient in study cohort. bAnalysis based on total of 31 study and nonstudy cGVHD patients with documentation of nail dystrophy. cAnalysis based on 11 cGVHD study patients. have preceding edema were on prednisone at time of cGVHD peripheral eosinophilia (eosinophil count 4500 cells/μL) than onset. Patients with sclerotic cGVHD were significantly more likely patients with nonsclerotic disease (12/14 versus 7/20, P = 0.005; to have peripheral edema than patients with nonsclerotic disease Table 2). Medical record review revealed no other identifiable (Table 2). causes of eosinophilia. In the subset of patients with eosinophilia (n = 19), the mean Control group. Of the 97 patients in our control population, time from the onset of eosinophilia to symptoms of cGVHD was peripheral edema was documented in 8 patients. Causes of 54 days (range 0–182 days). Eosinophilia persisted until diagnosis peripheral edema were identified in four of eight patients; two and treatment of cGVHD in all 19 patients. with arthritis and two with renal disease. The remaining four patients had self resolution of edema within 2 weeks, without Control group. Thirty-two patients of the control group had intervention. There was a significantly higher incidence of peripheral eosinophilia (4500 cells/μL) at some point during this peripheral edema in skin cGVHD patients compared with our time period. Of these 32 patients, 66% (21/32) had resolution of control group (Table 4). eosinophilia within 1 month or by the following visit. Cause of eosinophilia was identified in 20 patients, with most common Peripheral eosinophilia in skin cGVHD causes being eczema and infection. The mean peak peripheral eosinophil count and incidence of Thirty-four cGVHD patients (11 in study cohort; 23 in nonstudy eosinophilia was significantly higher in our cGVHD patients than in cohort) had documented eosinophil counts within 2 months our controls (Table 4). before the diagnosis. Mean peak peripheral eosinophil count was 1140 cells/μL (s.d. 1160). There was no significant difference in mean peak peripheral eosinophil count in patients with sclerotic DISCUSSION disease than in those with nonsclerotic disease. However, patients Nail dystrophy has been described in the literature as a with sclerotic cGVHD were significantly more likely to have manifestation of cGVHD.8 Many forms of nail dystrophy are a

© 2014 Macmillan Publishers Limited Bone Marrow Transplantation (2014) 1521 – 1527 Harbingers of severe skin cGVHD in children JT Huang et al 1526 result of injury or inflammation of the nail matrix. Disturbances of of cGVHD. Comparison of our findings with control patients the nail matrix, ordinarily a site of immune privilege, in GVHD may showed that nail dystrophy was uncommon in HSCT recipients at be an indication of severe immunologic dysfunction. However, our institution overall, and suggested that nail dystrophy may be a there has been no prior investigation of nail involvement as an significant and specific finding in this population. indicator of prognosis. Within our study cohort of 11 patients, PIU was present in all In our study, nail dystrophy was present in more than half of 5 patients with lung cGVHD, and was significantly associated with fi patients with skin cGVHD, the majority of whom had findings of poor pulmonary function. PIU is a speci c form of nail dystrophy longitudinal ridging, splitting and PIU. Our study demonstrated that is characterized by adherence of the distal portion of the fi nailbed to the ventral surface of the nail plate. PIU has been rarely nail dystrophy to be signi cantly associated with sclerotic 12,13 disease as well as lung cGVHD. In five patients with adequate reported as a manifestation of cGVHD. Interestingly, PIU has also been described as a characteristic nail finding in patients with documentation, nail dystrophy was present before the symptoms systemic scleroderma, but not in those with limited skin involvement (localized morphea).14 An association between PIU Table 4. Comparison of nail dystrophy, edema and eosinophilia and lung cGVHD has not been previously reported and the between skin cGVHD and control groups pathogenesis of this association is unknown. Thirty percent of control patients, and 14% of cGVHD patients a Characteristics Skin cGVHD Controls P-value had no documented nail exams in their medical record. Four of (36) (97) five patients with PIU in the study cohort had no prior o documentation of nail dystrophy, suggesting that the nail exam Nail dystrophy 0.0001 fi Yes 39% (14/36) 4% (5/97) may often be overlooked. Given our nail ndings, we recommend No 19% (17/36) 65% (63/97) careful evaluation of adnexal structures in HSCT recipients, paying Unknown/missing 14% (5/36) 30% (29/97) close attention to those with cGVHD. In patients with nail Peripheral edema o0.0001 dystrophy, particularly with findings of PIU, we suggest close Yes 44% (16/36) 8% (8/97) monitoring of PFTs to assess for lung cGVHD. No 47% (16/36) 92% (89/97) Peripheral edema as an early sign of sclerotic cGVHD has been Unknown/missing 8% (4/36) — previously reported.15–18 Recently, it has been postulated that o Mean peripheral eosinophil 1140 (1160) 480 (470) 0.0001 edema associated with cGVHD may be due to endothelial count in cells/μL (s.d.) (N = 34) (N = 97) 18 Eosinophiliab 0.025 activation and injury. In our study, peripheral edema was a Yes 57% (19/36) 33% (32/97) preceding sign in 13 of 15 patients with sclerotic cGVHD with a No 39% (15/36) 67% (65/97) mean time of onset of 58.5 days before the development of Unknown 4% (2/36) — sclerotic disease. In these patients, peripheral edema persisted If eosinophilia, self o0.0001 until sclerotic features became evident. Medical record review of resolution?c control patients showed that peripheral edema was not only Yes 0% (0/19) 66% (21/32) significantly less common in those without cGVHD, but also more No 100% 25% (9/32) likely to self-resolve in cases without identifiable cause. Although (19/19) other causes should be considered in these medically complex Unknown — 7% (2/32) patients, persistent peripheral edema in the post-HSCT recipient aP-values based on non-missing categories. Comparison of cGVHD versus should be evaluated with particular concern for sclerotic cGVHD. control for nail dystrophy and peripheral edema based on Fisher’s exact Peripheral eosinophilia has also been described as a preceding b test; P-value for eosinophil count based on two-sample t-test. Eosinophilia sign of cGVHD, but not a predictor of outcome, in both adults and fi 4 μ c fi de ned as peripheral eosinophil count 500 cells/ L De ned as children.10,19,20 In our study, patients with skin cGVHD had a resolution of eosinophilia within 1 month or at following lab draw. significantly higher mean peak peripheral eosinophil count than

Figure 4. Preceding edema was noted in 13 of 15 patients with sclerotic cGVHD.

Bone Marrow Transplantation (2014) 1521 – 1527 © 2014 Macmillan Publishers Limited Harbingers of severe skin cGVHD in children JT Huang et al 1527 control patients, with a mean time of onset of eosinophilia of 2 Mielcarek M, Martin PJ, Leisenring W, Flowers ME, Maloney DG, Sandmaier BM 54 days before the symptoms of cGVHD. Patients with sclerotic et al. Graft-versus-host disease after nonmyeloablative versus conventional cGVHD were more likely to have preceding eosinophilia than hematopoietic stem cell transplantation. Blood 2003; 25: 190–200. those with nonsclerotic disease. Although all cGVHD patients with 3 Baird K, Cooke K, Schultz KR. Chronic graft versus host disease (GVHD) in children. 57 – eosinophilia had persistently elevated eosinophil counts to the Pediatr Clin North Am 2010; :297 322. 4 Jacobsohn DA, Arora M, Klein JP, Hassebroek A, Flowers ME, Cutler CS et al. Risk time of diagnosis of cGVHD, 66% of control patients had factors associated with increased nonrelapse mortality and with poor overall resolution of eosinophilia at the following visit or blood draw. survival in children with chronic graft-versus-host disease. Blood 2011; 118: Our results corroborate previous data supporting peripheral 4472–4479. eosinophilia as an early marker for cGVHD, and suggest 5 Lee SJ, Klein JP, Barrett AJ, Ringden O, Antin JH, Cahn JY et al. Severity of chronic persistence of eosinophilia to be a more specific sign. graft-versus-host disease: association with treatment-related mortality and Finally, sclerotic cGVHD in adults has been associated with the relapse. Blood 2002; 100:406–414. need for prolonged immunosuppression, significant functional 6 Ito T, Ito N, Saathoff M, Stampachiacchiere B, Bettermann A, Bulfone-Paus S et al. impairment and, more variably, decreased survival.14,15 In our Immunology of the human nail apparatus: the nail matrix is a site of relative immune privilege. J Invest Dermatol 2005; 125: 1139–1148. study, patients with sclerotic disease also had worse outcomes, fi 7 Christoph T, Muller-Rover S, Audring J, Tobin DJ, Hermes B, Cotsarelis G et al. The with a signi cantly increased incidence of steroid-resistant disease human hair follicle immune system: cellular composition and immune privilege. and higher death rate. Patients with sclerotic disease were also Br J Dermatol 2000; 142:862–873. more likely to have nail dystrophy, preceding peripheral edema 8 Filipovich AH, Weisdorf D, Pavletic S, Socie G, Wingard JR, Lee SJ et al. National and preceding eosinophilia than those with nonsclerotic disease. Institutes of Health consensus development project on criteria for clinical trials in Although the temporal relationship between nail dystrophy and chronic graft-versus-host disease: I. Diagnosis and staging working group report. disease onset needs to be defined, our data suggests that these Biol Blood Marrow Transplant 2005; 11: 945–956. physical exams and laboratory findings may be harbingers of 9 Felder-Puig R, di Gallo A, Waldenmair M, Norden P, Winter A, Gadner H et al. Health-related quality of life of pediatric patients receiving allogeneic stem cell or severe skin cGVHD in children. bone narrow transplantation: results of a longitudinal multi-center study. Bone Given the rarity of chronic GVHD in the pediatric population, our Marrow Transplant 2006; 38:119–126. study was limited by small study size. In addition, our study cohort 10 Shulman HM, Sullivan KM, Weiden PL, McDonald GB, Striker GE, Sale GE et al. represented a severe subset of children with chronic GVHD, Chronic graft-versus-host syndrome in man. A long-term clinicopathologic study skewed largely by referral bias of treatment resistant patients to of 20 Seattle patients. Am J Med 1980; 69:204–217. dermatology. Thus, we incorporated retrospective data on cGVHD 11 Akpek G, Zahurak ML, Piantadosi S, Margolis J, Doherty J, Davidson R et al. patients who were not enrolled in our study whenever possible to Development of a prognostic model for grading chronic graft-versus-host 97 – further substantiate our findings. However, we were not able disease. Blood 2001; : 1219 1226. 12 Liddle BJ, Cowan MA. Lichen planus-like eruption and nail changes in a patient to use retrospective data to evaluate associations with PIU due to with graft-versus-host disease. Br J Dermatol 1990; 122: 841–843. limited available documentation. Finally, the cross-sectional nature 13 Palencia SI, Rodriguez-Peralto JL, Castano E, Vanaclocha F, Iglesias L. Lichenoid of our study limited our ability to establish a clear timeline of nail changes as sole external manifestation of graft vs host disease. Int J Dermatol events with regards to adnexal involvement and disease outcome. 2002; 41:44–45. Future longitudinal studies with pre- and post transplant 14 Patterson JW. Pterygium inversum unguis-like changes in scleroderma. Report of assessments should be useful in investigating the predictive value four cases. Arch Dermatol 1977; 113: 1429–1430. of our findings. 15 Schaffer JV, McNiff JM, Seropian S, Cooper DL, Bolognia JL. Lichen sclerosus This study investigated a group of pediatric skin cGVHD patients and eosinophilic fasciitis as manifestations of chronic graft-versus-host fi disease: expanding the sclerodermoid spectrum. J Am Acad Dermatol 2005; 53: at a single institution, and identi ed several cutaneous and 591–601. non-cutaneous features were associated with severe disease. Our 16 Janin A, Socie G, Devergie A. Fasciitis in chronic graft-versus-host disease. data suggest that nail dystrophy, persistent peripheral edema and A clinicopathologic study of 14 cases. Ann Intern Med 1994; 120: 993–998. persistent peripheral eosinophilia are harbingers of severe cGVHD 17 Kim KW, Yoon CH, Kay CS, Kim HJ, Lee EH, Park SY. Fasciitis after allogeneic of the skin in children. In addition, the presence of PIU may peripheral blood stem cell transplantation in a patient with chronic myelogenous be a harbinger of severe lung involvement in cGVHD. Larger leukemia. J Clin Rheumatol 2003; 9:33–35. prospective studies are indicated to confirm our findings. 18 Tardieu M, Rybojad M, Peffault de Latour R, Robin M, de Masson A, Xhaard A et al. Localized edema with sclerodermatous evolution: a possible form of skin chronic graft-versus-host disease associated with endothelial activation. Blood 2013; 122: CONFLICT OF INTEREST 463–465. 19 Ahmad I, Labbe AC, Chagnon M, Busque L, Cohen S, Kiss T et al. Incidence and fl The authors declare no con ict of interest. prognostic value of eosinophilia in chronic graft-versus-host disease after nonmyeloablative hematopoietic cell transplantation. Biol Blood Marrow Transplant 17 – REFERENCES 2011; :1673 1678. 20 Jacobsohn DA, Schechter T, Seshadri R, Thormann K, Duerst R, Kletzel M. 1 Leiper AD. Non-endocrine late complications of bone marrow transplantation in Eosinophilia correlates with the presence or development of chronic graft-versus- childhood: Part I. Br J Haematol 2002; 118:3–22. host disease in children. Transplantation 2004; 15: 1096–1100.

Supplementary Information accompanies this paper on Bone Marrow Transplantation website (http://www.nature.com/bmt)