Article Text

Download PDFPDF

Rosai-Dorfman disease: an overview
  1. Cassandra Bruce-Brand1,2,
  2. Johann W Schneider1,2,
  3. Pawel Schubert1,2
  1. 1 Stellenbosch University, Faculty of Medicine and Health Sciences, Division of Anatomical Pathology, Cape Town, South Africa
  2. 2 National Health Laboratory Service, Tygerberg Hospital, Cape Town, South Africa
  1. Correspondence to Dr Cassandra Bruce-Brand, Anatomical Pathology, Stellenbosch University, PO Box 241, Cape Town, 8000, South Africa; cassandra.bruce-bran{at}


Context Rosai-Dorfman disease is an uncommon histiocytic disorder most frequently presenting as bilateral cervical lymphadenopathy in children and young adults. Extranodal disease occurs in a significant proportion of patients. It has been recently classified as part of the ‘R group’ of histiocytoses by the Histiocyte Society in 2016. Cutaneous Rosai-Dorfman disease is regarded as a separate disease entity that falls into the ‘C group’ of histiocytoses according to this classification system. The pathogenesis was previously poorly understood; however, recent evidence demonstrating clonality in a subset of cases raises the possibility of a neoplastic process. A possible association with IgG4-related disease remains controversial.

Objectives To provide a comprehensive review of Rosai-Dorfman disease, including nodal, extranodal and cutaneous forms, with a particular emphasis on new insights into the possible clonal nature of the disease; to discuss the recently revised classification of the histiocytoses by the Histiocyte Society; and to summarise the findings from the literature regarding the controversial association with IgG4-related disease.

Data sources This review is based on published peer-reviewed English literature.

Conclusions Classic Rosai-Dorfman disease, which may be sporadic or familial, is considered a separate entity from cutaneous disease, which is reflected in the revised classification of histiocytoses. An increase in IgG4-positive plasma cells may be seen in Rosai-Dorfman disease. This finding in isolation is of limited significance and should be interpreted with caution. Studies investigating the molecular profile of the disease show that in at least a subset of cases the disease is a clonal process. The classification of Rosai-Dorfman disease is therefore likely to change as our understanding of the aetiopathogenesis evolves.

  • haematological disease
  • pathology, surgical
  • lymph nodes
View Full Text

Statistics from

Rosai-Dorfman-Destombes disease (RDD) is a rare histiocytic disorder described by Destombes in 1965 and later by Rosai and Dorfman in 1969 as ‘sinus histiocytosis with massive lymphadenopathy’ and previously classified by the Working Group of the Histiocyte Society of 1987 as a non-Langerhans cell (LC) histiocytosis.1–3 This same society has recently reclassified the histiocytoses based on new insights into the pathological, genetic and molecular features of these disorders. In this new classification, RDD now forms part of the ‘R group’ of histiocytoses, which includes familial RDD, sporadic RDD and other miscellaneous non-cutaneous, non-LC histiocytoses.4 Cutaneous RDD is classified separately as part of the ‘C group’ of histiocytoses.4 The diagnosis of RDD in the classic nodal form can usually be made with ease on routine H&E stained sections with a small panel of immunohistochemical markers. Extranodal forms or cases in which there is extensive fibrosis or scant emperipolesis may pose diagnostic challenges. Immunohistochemistry is an essential part of the work-up, as Langerhans cell histiocytosis (LCH) and other neoplastic histiocytoses must be excluded. Recent evidence shows some degree of overlap with IgG4-related disease, although this remains controversial.


The aetiopathogenesis of RDD is poorly understood. It has been previously perceived to be a reactive, non-neoplastic histiocytic disorder that lacks clonality and was therefore not included in the latest (2017) WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues. Evidence in support of a clonal nature of at least a subset of cases of RDD has recently come to light and is summarised here.

Kinase mutations have been recently described in nodal and extranodal (but not cutaneous) RDD, including mutations in ARAF,5 MAP2K1,6 7 NRAS 5 and KRAS.5 6 8–10 Up to 33% of cases of RDD showed KRAS or MAP2K1 mutations in one study, indicating that a subset of cases may indeed be clonal.6 A 2019 publication in which targeted DNA/RNA sequencing and whole exome sequencing were performed on 17 cases of RDD showed kinase driver mutations involving KRAS (4 of 17), MAP2K1 (2 of 17), NRAS (1 of 17), ARAF (1 of 17) and CSF1R (1 of 17).11 In addition, alterations in genes involved in intracellular trafficking (SNX24), transcriptional regulation (CIC, INTS2, SFR1, BRD4, PHOX2B), cell cycle regulation (PDS5A, MUC4), DNA mismatch repair (ERCC2, LATS2, BRCA1, ATM) and the ubiquitin proteasome pathway (USP35) were demonstrated.11

BRAF V600E mutations are described in histiocytic neoplasms such as LCH12 and Erdheim-Chester disease (ECD),13 14 and have been sought in RDD (see table 1). Ninety-one cases of RDD which show an absence of BRAF V600E mutations were found in the literature.13 15–18 Three cases of BRAF mutations have however been recently described.17 19 20 Fatobene et al 17 confirmed the presence of a BRAF V600E mutation in a single case of nodal RDD using multiplex picodroplet digital PCR. This mutation was not detected by pyrosequencing by the same authors.17 Mastropolo et al 20 also demonstrated a BRAF V600E mutation in systemic mixed RDD and LCH. Anti-BRAF V600E immunohistochemistry was strongly positive in both RDD and LCH, and BRAF p.V600E mutation was detected on peripheral blood mononuclear cells.20 Richardson et al 19 demonstrated a single somatic pathogenic mutation in exon 12 of the BRAF gene (p.486-491del) in a single central nervous system case of RDD. No data on the molecular profile of cutaneous RDD could be found. In the series of 21 cases published by Garces et al,6 the presence of KRAS and/or MAP2K1 mutations was associated with a younger age (p=0.0251), head and neck site (p=0.0009) and multifocal disease (p=0.0256).

Table 1

Summary of RDD cases assessed for BRAF mutations

Despite the expression of S100, the cell of origin is currently believed to be activated macrophages rather than dendritic cells.21 Various viral agents, such as human herpesvirus 6, parvovirus B19 and Epstein-Barr virus, have been investigated as potential inciting agents, but studies have failed to establish a causative link.22–24


The disorders recently classified as part of the ‘R group’ of histiocytoses are summarised in figure 1.4 Sporadic RDD is the most common form and includes the classic nodal form, extranodal RDD, neoplasia-associated RDD and immune disease-associated RDD.4 Neoplasms associated with RDD include lymphomas,25–42 leukaemias,43–45 malignant histiocytoses,46 47 LCH48–51 and ECD52 (see table 2). Systemic lupus erythematosus, idiopathic juvenile arthritis, autoimmune haemolytic anaemia and HIV are among the immune disorders that are associated with RDD.34 53

Figure 1

Schematic classification of Rosai-Dorfman disease according to the Histiocyte Society.4 AIHA, autoimmune haemolytic anaemia; ALPS, autoimmune lymphoproliferative syndrome; ECD, Erdheim-Chester disease; IJA, idiopathic juvenile arthritis; LCH, Langerhans cell histiocytosis; MH, malignant histiocytoses; NOS, not otherwise specified; SLE, systemic lupus erythematosus.

Table 2

Neoplasms associated with RDD

Familial RDD includes H syndrome (Faisalabad syndrome), an autosomal recessive genetic syndrome caused by mutations in the SLC29A3 gene, in which up to 20% of cases demonstrate nodal (19 of 79 cases) and/or extranodal (skin (23 of 79 cases) and nasal cavity (2 of 79 cases)) RDD.54–63 The syndrome is characterised by hyperpigmentation, hypertrichosis, hepatosplenomegaly, hearing loss, heart anomalies, hypogonadism, short stature (low height), hyperglycaemia and hallux valgus.62 Also included in familial RDD is FAS deficiency or autoimmune lymphoproliferative syndrome (ALPS)-related RDD caused by germline mutations in TNFRSF6.64 Forty-one per cent (18 of 44) of patients with confirmed ALPS were found to have nodal RDD in one study.64 Only one case of extranodal RDD, involving the spleen, has been described in this setting.47

Cutaneous RDD is classified under the ‘C group’ of histiocytoses.4 This group is further divided into xanthogranuloma and non-xanthogranuloma families (into which cutaneous RDD falls).4 Cutaneous RDD demonstrates unique epidemiological and clinical features and is therefore classified separately from other forms of RDD.65 Distinction from the xanthogranuloma family (S100-negative) is important and is discussed further in the Differential diagnosis section.

Clinical features

Classic RDD presents with massive bilateral painless cervical lymphadenopathy with associated fever, loss of weight and night sweats.2 66 It affects predominantly children and young adults with an average age of onset of 20.6 years, and occurs more commonly in African patients with a slight male predominance (male to female ratio of 1.4).34 Inguinal, retroperitoneal and mediastinal lymph nodes may also be involved.4 66 Extranodal disease is seen in over 40% of cases and may rarely occur in the absence of nodal disease, usually in older patients with different demographics.34 53 65 67 68 Common extranodal sites of involvement include the skin (10%),69 70 nasal cavity (11%),34 bone (5%–10%),34 71 orbital tissue (11%)34 and central nervous system (5%, predominantly dural).67 72–74 Bone involvement is seen as lytic lesions with well-defined sclerotic margins on radiography and is associated with nodal disease in 10% of cases.34 71 75 Central nervous system involvement may mimic meningioma clinically and is usually not associated with nodal disease.67 74 An elevated erythrocyte sedimentation rate, leucocytosis, hypergammaglobulinaemia and autoimmune haemolytic anaemia may be observed on laboratory studies.53

Cutaneous RDD presents in patients with an older mean age (43.5 years) than nodal RDD, a higher proportion of female patients (2:1 ratio) and a higher proportion of Asian and white patients.65 These patients typically lack an association with systemic or extracutaneous disease, and the disease process tends to remain localised despite long-term follow-up.65 68 In a series of 25 cases from China, the majority of patients (80%) presented with papulonodular lesions.68 Other presentations described include indurated plaques, tumour-like lesions, acneiform lesions and eruptive xanthoma-like lesions.68 76 77

Pathology findings

Gross features

Lymph nodes involved by RDD are enlarged and matted together forming firm multinodular masses with a yellow-white appearance on cut section.2 34 Fibrosis of the capsule may be evident.2

Microscopic features

The striking histological finding in RDD is an enlarged node with a low power appearance of extensive sinusoidal expansion (figure 2).2 In more advanced cases there is effacement of the nodal architecture with a diffuse infiltrate of histiocytes. The cortex comprises numerous activated B cells and mature plasma cells with few follicles which, together with the pale histiocytes, give the appearance of alternating dark and light zones. The sinusoids contain numerous large histiocytic cells with smooth contoured hypochromatic nuclei, small distinct centrally placed round nucleoli and ill-defined, pale, wispy cytoplasm (figure 3).2 34 Multinucleation, atypia and rare mitotic figures may be seen; however, the majority of cells usually retain smooth nuclear contours and abundant pale cytoplasm. Emperipolesis, a useful but not specific feature, is seen as intact haematolymphoid cells within a vacuole or floating freely in the cytoplasm of the histiocytes (figure 4). Neutrophils may be seen in the background (figure 5), occasionally forming micro-abscesses, and eosinophils are typically absent. Marked sclerosis with a storiform architecture and lobulation may be seen (figure 6). Extranodal disease mimics nodal RDD closely but may display more prominent lymphoid follicles with germinal centres, fibrosis, sclerosis, fewer histiocytes and more subtle emperipolesis (figure 7). Emperipolesis is not a requirement for the diagnosis and is often inconspicuous at extranodal sites in particular.69 A careful search for coexisting pathology in nodal disease is required as RDD can occur with lymphoma, ECD and LCH, as described above. In such cases the specimen should demonstrate RDD in more than 10% of the tissue to constitute a diagnosis of neoplasia-associated RDD.69

Figure 2

H&E 40×. Low power view of nodal Rosai-Dorfman disease showing alternating light and dark areas. The dark areas comprise an admixture of lymphocytes and plasma cells. The light areas comprise the characteristic pale pink histiocytic cells of Rosai-Dorfman disease.

Figure 3

H&E 200×. Nodal Rosai-Dorfman disease showing sheets of large pale pink cells with wispy cytoplasm and vesicular nuclei.

Figure 4

H&E 200×. Nodal Rosai-Dorfman disease showing characteristic cells with abundant wispy cytoplasm and admixed inflammatory cells including lymphocytes and plasma cells. Emperipolesis is seen.

Figure 5

H&E 200×. Nodal Rosai-Dorfman disease showing characteristic cells with emperipolesis of other inflammatory cells. Neutrophils are evident in this image.

Figure 6

H&E 100×. Nodal Rosai-Dorfman disease showing marked fibrosis which may be storiform.

Figure 7

H&E 20×. Extranodal Rosai-Dorfman disease showing sheets of pale histiocytes with scattered lymphoid aggregates and follicles.

Cutaneous RDD is typically a dermal-based lesion, sometimes involving the dermis and subcutaneous tissue (figure 8).65 The overlying epidermis may show acanthosis, increased basal keratinocyte pigmentation, ulceration or an epidermal collarette. The lesion is typically nodular and poorly circumscribed with infiltration into adjacent tissue. The typical histiocytic cells with emperipolesis as described above are characteristic. These cells may form bands or sheets giving a light and dark zonal appearance as described above or may be haphazardly distributed with more of a starry-sky low power appearance (figure 9). Plasma cells and neutrophils are often present with formation of micro-abscesses.

Figure 8

H&E 20×. Cutaneous Rosai-Dorfman disease showing expansion of the dermis by an infiltrate of large pale cells admixed with lymphocytes.

Figure 9

H&E 100×. Cutaneous Rosai-Dorfman disease showing sheets of pale cells with bands of lymphocytes giving a light and dark zonal appearance.

Features seen on fine needle aspiration

Fine needle aspirates show a heterogeneous population of lymphoid cells, including lymphocytes, plasma cells and scattered large histiocytes with oval vesicular nuclei and conspicuous nucleoli.78 The cytoplasm of the histiocytes is abundant and wispy and emperipolesis may be seen (figure 10).

Figure 10

Giemsa stain 1000× under oil. Large histiocyte with emperipolesis in an aspirate of a cervical lymph node.

Ancillary studies

The histiocytes are S100, CD68 and CD163 positive and are by definition CD1a and langerin (CD207) negative, thereby excluding LCH. The S100 stain often highlights the emperipolesis (figure 11). The plasma cell population in the cortex will stain with plasma cell markers (CD38, CD138 and MUM1) and may demonstrate abundant IgG4-positive plasma cells (figure 12).79 Based on consensus expert opinion, the Histiocyte Society recommends that all cases of RDD be evaluated for IgG4-positive plasma cells (grade D2 evidence) (see the the Differential diagnosis section for further discussion on this issue).4

Figure 11

S100 protein immunohistochemical stain 200×. The S100 highlights the histiocytes as well as the emperipolesis (lack of staining of the inflammatory cells within the cytoplasm of the histiocytes).

Figure 12

IgG4 immunohistochemical stain 200×. Numerous plasma cells staining positive for IgG4 are highlighted in this case of Rosai-Dorfman disease.

Differential diagnosis

Perhaps the most important differential diagnosis of RDD is that of LCH. A diagnosis of RDD by definition requires the exclusion of LCH by negative CD1a or CD207 staining of the histiocytic infiltrate. Other morphological clues to this differential are the absence of an eosinophilic infiltrate in RDD, the characteristic elongated grooved nuclear features of LCH and the absence of a prominent plasma cell component in LCH. Of note, S100 positivity is observed in both RDD and LCH. RDD may be associated with LCH as discussed previously. Two distinct populations of cells with different immunohistochemical staining (one CD1a-positive, ±S100; and one CD1a-negative, +S100) would need to be demonstrated to make this diagnosis.

Nodal RDD may also simulate other conditions with a prominent sinusoidal pattern, such as sinus histiocytosis, anaplastic large cell lymphoma (ALCL), metastatic carcinoma and malignant melanoma. This can usually be resolved with immunohistochemistry and close attention to the H&E morphology. Sinus histiocytosis lacks the characteristics described above of the typical RDD histiocytic cells and would be negative for S100. ALCL will demonstrate more nuclear pleomorphism with typical hallmark cells which will be CD30-positive. Metastatic carcinoma and melanoma will not usually show emperipolesis and carcinoma will be positive with pan-cytokeratin markers. Melanoma will be S100-positive but will also stain positive with HMB45, Melan-A and SOX10. Other disorders such as Gaucher disease, Whipple disease and Hodgkin’s lymphoma may enter the differential diagnosis in nodal disease. Gaucher disease demonstrates histiocytes filled with finely fibrillar, tissue paper-like cytoplasm due to accumulation of sphingolipids. The nuclei typically have indistinct nucleoli and lack emperipolesis. In Whipple disease the macrophages are packed with periodic acid-schiff positive diastase-resistant bacilli and nodal involvement is typically mesenteric. Classic Hodgkin’s lymphoma will usually show more typical Reed-Sternberg or Hodgkin’s cells with macro-nucleoli, lacks S100 positivity, and the large cells stain positive with CD30 and CD15 with a characteristic Golgi pattern of staining.

Within the head and neck region rhinoscleroma, granulomatosis with polyangiitis and extranodal natural killer/T cell lymphoma may be considered. Again, close attention to the H&E morphology with positive S100 staining of the histiocytes can resolve most of these differentials.

Cutaneous involvement must be distinguished from juvenile xanthogranuloma (JXG). JXG is characterised by macrophages and spindled cells with scattered Touton-type multinucleated giant cells, negativity for S100 and lack of emperipolesis. The storiform sclerosis and presence of plasma cells that may be seen, particularly in cutaneous RDD, raise the differential diagnosis of IgG4-related sclerosing disease, which is discussed further in the next section.

The link to IgG4-related disease

RDD may share some morphological features with IgG4-related disease, such as storiform fibrosis and abundant plasma cells, thereby raising this condition as a possible differential diagnosis. This is particularly true in cutaneous disease, which often shows marked sclerosis with a storiform architecture. In 2009 Kuo et al 79 documented a possible association between RDD and IgG4-related disease by demonstrating high levels of IgG4-positive plasma cells in cases of cutaneous RDD. Of the 12 cutaneous cases assessed in this study, 11 showed more than 30 IgG4-positive cells per high power field (HPF), with a mean IgG4 to IgG ratio of 34% (see table 3).79 This was followed by other studies which showed similar findings of abundant IgG4-positive plasma cells with raised IgG4 to IgG ratios in RDD at nodal and extranodal sites (table 3).79–82 Two pulmonary cases of RDD are also reported to have shown increased IgG4-positive plasma cells; however, the exact number of positive cells and the IgG4 to IgG ratio were not published and were scored as ‘significant’83 and ‘score of 2’ (11–30 IgG4-positive plasma cells/HPF).84 These findings were not corroborated in other studies of RDD and the association between these disorders has not been widely accepted.85 Liu et al 86 described only 3 of 29 cases with IgG4 to IgG ratio of more than 40%. Richter et al 87 and Chen and Lee88 did not find an increase in IgG4-positive plasma cells in one parotid and one cardiac case.

Table 3

Summary of findings of increased IgG4-positive plasma cells in RDD in the literature

In 2012 a consensus statement on the pathology of IgG4-related disease was published.85 Histological criteria for the diagnosis of IgG4-related disease (storiform fibrosis, obliterative phlebitis and a dense lymphoplasmacytic infiltrate) with specific cut-off points for the number of IgG4-positive plasma cells by site were defined. A three-tiered diagnostic terminology based on pathological features was proposed, including ‘histologically highly suggestive of IgG4-related disease’, ‘probable histological features of IgG4-related disease’ and ‘insufficient histopathological evidence of IgG4-related disease’.85 The consensus statement concludes with minimal criteria required to propose a new diagnosis of IgG4-related disease, which includes the histological criteria as well as raised serum IgG4 levels, response to glucocorticoid therapy, and other organ involvement consistent with IgG4-related disease.85 A diagnosis of IgG4-related disease thus requires clinical, serological and histological evidence and not one of these in isolation is sufficient. They further state that interpretation of isolated increased IgG4-positive plasma cells should be made with caution as this is a relatively non-specific finding that can be seen in many inflammatory conditions.85

In 2015 an expert panel published a consensus guidance statement on the management and treatment of IgG4-related disease in which they state that the most accurate assessment of IgG4-related disease is based on full clinical history, physical examination, selected laboratory investigations and appropriate radiology studies.89 The panel states that neither clinical or pathological features alone are sufficient to make a diagnosis and that disorders that mimic IgG4-related disease must be rigorously excluded.89 RDD is listed as one of the disorders that may mimic IgG4-related disease clinically or histologically.

It is clear from these guidelines that an increased number of IgG4-positive plasma cells in isolation has no diagnostic utility and that IgG4 immunohistochemical staining should be performed and interpreted with great caution. It has however been recommended by the Histiocyte Society in the revised classification of the histiocytoses that all cases of RDD should be evaluated for IgG4-positive plasma cell infiltration (based on expert opinion: grade D2 evidence).4 No further guidelines are provided on how these results should be interpreted, and to our knowledge no such guidelines within the context of RDD currently exist.

In our practice, based on the recommendation of the Histiocyte Society, an IgG4 immunohistochemical stain is performed on all cases. Three ×40 fields (22 field number eyepiece) with the highest number of positive staining cells are counted and an average calculated as suggested in the consensus statement.85 The result is documented as the number of IgG4-positive plasma cells per HPF. Further information regarding this calculation can be seen in the consensus guidelines.85 Based on the clear guidelines provided for the diagnosis of IgG4-related disease, a comment is added that the significance of this result in isolation without other clinical, serological or radiological evidence for IgG4-related disease is uncertain and that isolated increased numbers of IgG4-positive plasma cells may be seen in RDD.

Current treatment and prognosis

Sporadic RDD is usually self-limited and has a good outcome, with spontaneous remission reported in up to 50% of cases.90 Up to 10% of patients may die of their disease due to direct complications, infections and amyloidosis.34 90 Consensus recommendations for the work-up and management of RDD were published in 2018 and are beyond the scope of this review.69 Briefly summarised here, observation is indicated in patients with uncomplicated adenopathy and asymptomatic cutaneous disease.69 Surgical excision may be indicated in unifocal extranodal disease or for symptomatic airway, cranial, spinal or sinus disease.69 Those with multifocal irresectable extranodal disease may require systemic therapy of which there is currently no standardised regimen. Systemic therapies include corticosteroids, sirolimus, radiotherapy, chemotherapy and immunomodulatory therapy.69 Sufficient evidence is currently lacking to support firm associations between prognosis and underlying molecular alterations. The consensus recommendations advise targeted next generation sequencing for Mitogen-Activated Protein Kinase (MAPK) mutations in severe or refractory disease with consideration of targeted therapy if driver mutations are identified.69


Classic RDD is classified as part of the ‘R group’ of histiocytoses. Cutaneous RDD is considered distinct with different epidemiology and clinical features and falls into the ‘C group’ of histiocytoses. The diagnosis can be made by recognising the characteristic S100-positive histiocytes with emperipolesis that expand the sinusoids and impart a dark and light zonal appearance to the tissue. Diagnostic pitfalls include the presence of extensive fibrosis is some cases (particularly in cutaneous RDD), scant emperipolesis in extranodal disease and failure to exclude important differential diagnoses. It is currently recommended that all cases be evaluated for IgG4-positive plasma cells. Any reported increase in IgG4-positive plasma cells should be accompanied by a comment that this finding requires interpretation in conjunction with an appropriate clinical, serological and radiological context. Our understanding of the aetiopathogenesis of RDD is evolving. Although not currently classified as a neoplastic disorder, recent studies showing clonality in some cases have changed our understanding of the disease. Further characterisation of the spectrum of molecular alterations in RDD is required. The implications of these findings for prognosis and treatment purposes remain to be determined.

Take home messages

  • RDD is classified as part of the ‘R group’ of histiocytoses and includes classic nodal disease, extranodal disease, neoplasia associated and autoimmune associated disease. It may be sporadic or familial.

  • Cutaneous RDD is classified separately as part of the ‘C group’ of histiocytoses and must be distinguished from juvenile xanthogranuloma.

  • At least a subset of cases are clonal with mutations involving MAP2K1, KRAS, NRAS, ARAF and CSF1R. BRAF mutations are uncommon compared to Langerhans cell histiocytosis and Erdheim-Chester disease.

  • An increase in IgG4 positive plasma cells should be documented but interpreted in context


View Abstract


  • Handling editor Runjan Chetty.

  • Contributors All authors have contributed to the production of this manuscript.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.