Objective To analyse the clinicopathological presentation, outcome and importance of bone marrow haemophagocytosis in patients with infection-associated haemophagocytic lymphohistiocytosis (IA-HLH) in a tertiary care hospital in Northern India.
Study design Between January 2007 and December 2009, 26 consecutive patients meeting the diagnostic criteria for IA-HLH, based on the HLH2004 protocol of the Histiocyte Society, were followed up for between 12 and 34 months (median 20 months).
Results IA-HLH was diagnosed in three of the five patients who died 5–6 weeks after the onset of the illness, whereas diagnosis in the remaining group was made a median of 2 weeks after the onset of the illness. The predominant presenting features were fever (100%), hepatomegaly (69%), splenomegaly (58%) and anaemia (96%). All patients showed >3% haemophagocytosis on bone marrow studies—in four cases after serial aspiration/biopsies. Twenty-one (80.8%) cases were non-fatal and five (19.2%) patients died. The non-fatal cases included eight (38.1%) cases of viral infection, seven (33.3%) bacterial infections, two (9.6%) fungal and four (19.0%) protozoal infections; whereas four (80%) bacterial infections and one (20%) viral infection were associated with the fatal cases. The mean of the nadir blood counts of white blood cells, absolute neutrophil counts and platelets; the mean of all the peak biochemical parameters of liver function tests, lactate dehydrogenase and ferritin and the lowest fibrinogen values before treatment, differed significantly (p<0.05) between the non-fatal and the fatal group, being worse in the latter.
Conclusions IA-HLH is important because it can obscure the typical clinical features of the underlying primary disease, thus delaying the diagnosis and having a negative effect on the outcome. Although bone marrow haemophagocytosis is not a mandatory diagnostic criterion, we found it to be a useful tool together with biochemical parameters for early recognition of HLH, especially in developing countries lacking molecular and flow laboratories. The severity of pancytopenia and derangement in biochemical markers were significantly higher in the patients who died.
Statistics from Altmetric.com
Haemophagocytic lymphohistiocytosis (HLH) was first described in 1939 as histiocytic medullary reticulosis.1 HLH has been described as a heterogeneous group of disorders presenting with high fever, hepatosplenomegaly, lymphadenopathy, typically combined with variable cytopenias, haemophagocytosis, hyperferritinaemia, hypercytokinaemia, hypertriglyceridaemia coagulopathy and hypofibrinogenaemia. There are two distinct forms of HLH, classified as primary and secondary HLH. Primary HLH is an autosomal recessive disorder, also termed familial haemophagocytic lymphohistiocytosis. It commonly occurs in the early years of life, but is also known to present later, being fatal when untreated.2 Defects in a number of genes have been linked to familial haemophagocytic lymphohistiocytosis—for example, Perforin (PRF1), Munc 13-4 (UNC13D), Syntaxin 11 (STX11), Munc 19-2 (STXBP2). Secondary HLH is triggered by a variety of diseases such as infections, immunodeficiency syndromes, haematological neoplasias and autoimmune diseases. This classification is mostly artificial as primary HLH can present at any age, and both types could be triggered by a variety of infections compounding the diagnostic dilemma.2 This classification is useful in long-term planning of treatment, as mortality is universally high in primary HLH unless treated by haematopoietic stem cell transplant.
Infection-associated HLH (IA-HLH) is the commonest form of secondary HLH. IA-HLH was originally described in viral diseases by Risdall et al in 1979.3 Viral infections are the commonest trigger and include Epstein–Barr virus (EBV), cytomegalovirus, HIV, human herpes virus 8 (HHV8), parvovirus, influenza virus, hepatitis virus and enterovirus, the most common being EBV.3–5 Rarely, dengue virus has been reported as a trigger in IA-HLH. However, the frequency of its association was higher in our study as was also reported previously from India by Ramachandran et al.6 Since the initial description, IA-HLH has also been documented in patients with bacterial, parasitic or fungal infections. Mortality is affected by the severity and nature of the triggering event, and the delay in establishment of diagnosis.2 Various investigators have diligently and aggressively searched for clinical, biochemical and haematological parameters to identify markers to facilitate early diagnosis, and to stratify these patients according to risk and determine the prognosis. A number of treatments, including immunomodulation, together with specific treatment of the underlying disorder with appropriate supportive care have been tried with varying results.
The pathophysiology of HLH remains to be elucidated for acquired forms; the genetic forms are better understood. HLH is characterised by multisystem inflammation and is a reactive process as a result of prolonged and excessive activation of antigen-presenting cells (APCs) (includes macrophages and histiocytes) and CD8 T cells. These activated APCs lead to upgrading of CD163, which is the haem scavenging receptor through which haemophagocytosis is mediated. Most of the known genetic causes lead to cytotoxic dysfunction of natural killer (NK) and T cells, thus obstructing the immunological ‘switch-off’ mechanisms. Persistent activation of APCs and T cells leads to hypercytokinaemia with increased levels of interleukin (IL)-1B, tumour necrosis factor α, IL-6 and IL-8. Raised levels of interferon λ are commonly seen in EBV-associated IA-HLH.7 ,8
Tropical infections and EBV-associated HLH (including EBV-related malignancies) are more common in Asian populations.9 Surprisingly, data on IA-HLH secondary to tropical infections are sparse from India.10 In view of this, a study was conducted to analyse the clinicopathological presentation and outcome of patients with IA-HLH in a tertiary care hospital in Northern India.
Army Hospital Research and Referral, New Delhi, is the apex tertiary care institution of the Indian armed forces medical services with a large catchment area covering almost all of India.
Patients and methods
Diagnosis and monitoring
Twenty-six patients diagnosed with IA-HLH between January 2007 to December 2009, who met the diagnostic criteria (box 1) laid down by the study group of the Histiocyte Society 2004 were enrolled and analysed.11 ,12
Diagnostic criteria modified from study group of the Histiocyte Society 2004.
Fever: peak temperature >38.5°C for ≥7 days
Splenomegaly: spleen palpated >3 cm below the left costal margin
Cytopenia involving two or more cell lines: haemoglobin <9.0 g/dl, or platelets <100 000/µl, or absolute neutrophil count <1000/µl
Hypertriglyceridaemia or hypofibrinogenaemia: fasting triglycerides >2.0 mmol/l (177 mg/dl), or >3 SD above the normal value for age, or fibrinogen <1.5 g/l, or >3 SD below the normal value for age
Haemophagocytosis: demonstrated in bone marrow, spleen, or lymph node. No evidence for malignancy
Low or absent natural killer cell activity
Serum ferritin level >500 µg/l
Soluble CD25 (soluble interleukin 2 receptor) >2400 U/ml
The diagnosis of haemophagocytic lymphohistiocytosis (HLH) requires the presence of all five major criteria. Either criterion (A) or a combination of criteria (B) and (C) may be a substitute for one of the major criteria. If a patient meets only four criteria and the clinical suspicion for HLH is high, appropriate treatment must be started, as delays may be fatal. In the appropriate clinical setting, the diagnosis is justified by a positive family history of HLH; parental consanguinity is only suggestive of HLH.
All patients were managed as inpatients either in acute ward or in an intensive care unit and received supportive care with specific antimicrobial treatment as indicated. Complete blood count, biochemical parameters and electrolytes were serially monitored. Once HLH was suspected, serial serum ferritin, lactate dehydrogenase (LDH) and fibrinogen were measured. In all cases with a high ‘index of clinical suspicion’, bone marrow aspiration and biopsy were carried out from the posterior superior iliac spine using a Jamshidi needle to demonstrate haemophagocytosis (figures 1 and 2). Serial bone marrow aspiration and biopsy were invaluable in establishing the diagnosis. In four cases, the initial bone marrow examinations were inconclusive or did not show haemophagocytosis in >3% of cells, despite pancytopenia on peripheral smear, with a suggestive clinical picture. These patients were followed up with biweekly bone marrow studies and were found to have significant haemophagocytosis (>3%) on subsequent bone marrow smears. As a policy in our centre, bone marrow aspiration is combined with bone marrow biopsy when investigating pancytopenia, as they complement each other, morphology by aspiration and cellularity, granulomas and deposits, being better demonstrated by biopsy.
All 26 patients underwent an autoimmune investigation in addition to a thorough search for the focus of infection. Enteric fever was diagnosed on the basis of clinical features, positive blood culture for Salmonella typhi and rising Widal titres. Kala azar was diagnosed when bone marrow showed the presence of Leishmania donovani bodies or a positive PCR for leishmania was obtained (in one case where Leishmania donovani bodies were not seen on bone marrow examination). Dengue infection was diagnosed on the basis of clinical syndrome, positivity for NS1 antigen and presence of IgM antibodies, with dengue PCR being carried out only to establish the start of an epidemic. EBV infection was diagnosed with EBV–PCR and the presence of IgM antibodies to capsid antigen. The diagnosis of malaria was made by immunochromatographic tests and demonstration of parasite in blood smear and invasive aspergillosis on tissue biopsy. Subacute bacterial endocarditis was diagnosed from the clinical features and demonstration of vegetation on transoesophageal echocardiography. Adult onset Still's disease and Kikuchi's disease were diagnosed on the basis of clinical presentation, exclusion of other diseases, demonstration of markedly raised serum ferritin and necrotising lymphadenopathy, respectively. Those cases with short febrile illness and self-limiting pancytopenia with no definite underlying cause, were labelled as ‘unspecified viral infection’. The bone marrow was cellular and reactive in all these cases, with evidence of >3% haemophagocytosis.
The immediate aim of treatment is suppression of the exaggerated inflammatory response using immunomodulatory/immunosuppressive agents. This has to be complemented by aggressive supportive care and treatment of the underlying infection with appropriate antimicrobial agents. Management of IA-HLH remains largely supportive with treatment of the underlying infection. The cornerstones of management of IA-HLH included:
specific antimicrobial agents;
Specific antimicrobial agents were given according to the hospital infection committee protocol to all those patients in whom the infective agent could be isolated. Immunomodulatory treatment with steroids/intravenous immunoglobulin (IVIg) was added where response was suboptimal, and to those with unidentified infection. However, all patients with febrile neutropenia received empirical antibiotics with supportive care. In neutropenic patients with fever, empirical initiation of antibiotics, included a third-generation antipseudomonal cephalosporin and an aminoglycoside. Antistaphylococcal glycopeptide was added when fever persisted for >48 h despite treatment. If febrile neutropenia persisted beyond 5–7 days, antifungal agents were started. Enteric fever was managed with ceftriaxone 2 g given intravenously every 12 h. Visceral leishmaniasis was treated with liposomal amphotericin (total dose 20 mg/kg). Severe malaria was managed with artemesinin combination therapy (artesunate and doxycycline). Tuberculosis was treated with the antitubercular standard 4 drug (EHRZ). Antifungal treatment included liposomal amphotericin/capsofungin for candidaemia and voriconazole for invasive aspergillosis.
Immunomodulatory treatment with IVIg with or without steroids was started in patients who had a suboptimal response to appropriate antimicrobial agents with supportive care, as shown by deterioration in clinical status and worsening of pancytopenia. Parenteral steroids were given as dexamethasone/methylprednisolone pulses. IVIg when administered, was in doses of 400 mg/kg for 5 days. Treatment of patients with EBV infection included etoposide in addition to steroids and IVIg. Ciclosporin in a dose of 1.5–3 mg/kg infusion in two divided doses, to achieve target levels of 200–250 ng/ml, was also given to the patient with Kikuchi's disease with EBV infection. Granulocyte-colony stimulating factor (G-CSF) was given to three patients for 3–5 days in peripheral hospitals, as part of their treatment before their referral to our hospital. After G-CSF administration, all these three patients deteriorated clinically with worsening of their pancytopenia resulting in death.
Supportive care included the following:
Blood component treatment.
Packed red blood cells: leuco-depleted and irradiated to maintain haemoglobin (Hb)≥9 g/dl
Platelet transfusion. Prophylactic transfusion was avoided. However, all patients with thrombocytopenia and bleeding were transfused with group-specific, irradiated single-donor platelets harvested on Cobe Spectra/Haemonetics MCS-3p cell separators. If single-donor platelets were not available, 6–8 units of random donor platelets were given instead.
In the presence of coagulopathy/disseminated intravascular coagulopathy, fresh frozen plasma 10 ml/kg and cryoprecipitate, 1 unit/10 kg body weight (when fibrinogen levels<150 mg/ml) were transfused.
Nutrition: enteral/total parenteral nutrition was administered when indicated as these patients were universally hypercatabolic, presumably because of underlying hypercytokinaemia and poor oral intake.
Elective ventilation/lung protective ventilation in cases of sepsis with acute respiratory distress syndrome.
Fluid and electrolyte management with close monitoring.
The 26 patients were categorised into two groups in which the disease was non-fatal or fatal. Difference between means was assessed by the non-parametric Mann–Whitney U test for continuous data. Proportions were calculated for dichotomous data. Data were analysed using SPSS V.17. p Values < 0.05 were regarded as significant.
Patient characteristics are summarised in tables 1⇓–3. Of the 26 patients, 18 were male and eight were female. The age group was 9–75 years, mean age 33.7 years. No significance was noted in gender and age distribution. All patients had fever with a duration of 2–8 weeks (median 3 weeks). The time from onset of the disease to diagnosis was 1–6 weeks (median 2 weeks). Family history of IA-HLH was negative in all patients. The 21 non-fatal cases were associated with various underlying infections which included eight (38.1%) cases of viral infection, seven (33.3%) bacterial infections, two (9.6%) fungal and four (19.0%) protozoal infections (table 2), whereas, four (80%) bacterial infections and one (20%) viral infection were associated with the fatal cases. Hepatomegaly was seen in 18/26 (69%) patients, splenomegaly in 15/26 (58%) and lymphadenopathy in 7/26 (27%) cases. Rash was noted in 9/26 (35%) cases and neurological features were seen in only one case (4%). All patients had haemophagocytosis as demonstrated by bone marrow studies. Three patients had malignancy in remission (acute lymphoblastic leukaemia, acute myeloid leukaemia and non-Hodgkin's lymphoma (NHL)) and two had associated stable autoimmune diseases (adult-onset Still's disease and Kikuchi's disease). The presence of comorbid diseases had an adverse effect on outcome. Three of 21 (14.3%) non-fatal cases and 2/5 (40%) fatal cases had associated comorbid conditions—namely, adult-onset Still's disease and NHL. Comorbidities in the non-fatal cases were acute lymphoblastic leukaemia, acute myeloid leukaemia and Kikuchi's disease. Three patients who were given G-CSF had worsening of their febrile syndrome with tachycardia, tachypnoea, increasing rashes, falling blood counts and raised transaminases, ending in death.
Serial blood counts and biochemical parameters were of utmost importance for the early identification of IA-HLH. Interestingly, it was noted in our study, that when the ferritin levels peaked, cytopenias were at their lowest point, and the patients’ clinical condition was at its worst. The mean of the nadir white blood cell counts, absolute neutrophil counts and platelets; the mean of all the peak biochemical parameters of liver function tests, LDH and ferritin and the lowest fibrinogen values before treatment differed significantly (p<0.05) between the non-fatal and the fatal group (tables 4 and 5), being worse in the latter. The severity of pancytopenia was significantly worse in the fatal group. The blood counts included the mean values of the lowest white blood cell counts (771.4 vs 200.0, p<0.001), absolute neutrophil counts (354.3 vs 64.0, p<0.001) and platelets (32.5 vs 7.2 (×103, p<0.005)) (table 4). There was no significant difference in peak levels of triglyceride between the fatal and non-fatal groups (p=0.179).
In our hospital we lacked the facility to estimate NK cell activity and concentrations of soluble CD25 receptor. In addition, we had no facilities for genetic and mutational analysis for molecular diagnosis.
Twenty-one of the 26 patients became well and healthy with a median follow-up of 20 months, ranging from 12 to 34 months. Five (19.2%) patients died during the course of treatment. A 9-year-old girl with severe bacterial pneumonia died within 24 h of her diagnosis. One patient with enteric fever developed septicaemic shock with disseminated intravascular coagulopathy and died, despite energetic treatment. Three patients who were given G-CSF, developed worsening of febrile syndrome and subsequently died. They included a 33-year-old man with EBV-associated HLH, who deteriorated rapidly and had persistent severe pancytopenia, developed bilateral pneumonia and died as a consequence; a 32-year-old man, known to have adult-onset Still's disease, receiving maintenance treatment with immunosuppressive agents, who presented with bacterial pneumonia and encephalopathy and died after recurrent seizures; a patient with NHL who developed febrile neutropenia and died of polymicrobial (candida and pseudomonas) sepsis.
Twenty-three patients had documented infection and three had non-specified viral fever as the trigger for HLH. We presume that the 21 non-fatal cases belong to the secondary HLH group as there was a negative family history and no subsequent episodes of infection/recurrence of HLH on a median follow-up of 20 months (12–34 months). However, it is not possible to presume the same for the patients who died, as underlying primary HLH might have been missed and this is recognised as a limitation of the study. In contrast to patients in Western countries, the familial forms of HLH, though found in Asian patients, are uncommon.8 ,13 Ramchandran et al6 reported a series of 33 children from India who met the HLH 2004 criteria for diagnosis, where the overall survival rate was noted to be 76%. Genetic analysis, determination of NK cell function and soluble CD25 receptor assay, however, were not carried out in their study. The high survival rates were ascribed to early diagnosis and early institution of immunomodulatory treatment and possibly, high incidence of secondary HLH. In their study, 14/33 (42%) patients had an infection, with 5/14 (36%) infections due to dengue virus and 3/14 (21%) due to EBV. In our study, we found dengue virus in 4/26 (15.4%) cases and one case of EBV as an infective trigger. The overall survival rate in our study was 21/26 (80.8%).
Among viral infections, EBV is undoubtedly the major cause of HLH, as cited in various reports. Published data have shown that very high levels of proinflammatory cytokines are associated with EBV-related HLH among Asians.14 Apart from EBV, dengue infection, especially in epidemics, emerged as an important cause of IA-HLH in our study, which was fortunately self-limiting and not associated with any death. The same has been reported by Ramchandran et al6 from Chennai, India.
Bacterial infections also form a major portion of IA-HLH in this part of the world. These patients have a relatively good prognosis if timely antibiotics and supportive measures are instituted. Tuberculosis, enteric fever, ehrlichia and other bacterial infections, have been detected in patients with HLH.2 In our study, enteric fever was noted to be the leading cause among bacterial infections.
Among other infections, both falciparum and vivax forms of malaria and leishmaniasis were associated with HLH. A review of literature also shows an association of toxoplasma, babesiosis and strongyloidiasis with HLH.15 Fungal infections, especially disseminated, when the host is immunocompromised have been shown to cause HLH.16 ,17 We had three cases of fungal infections associated with HLH and another with a suspected fungal infection superimposed on tuberculosis, which responded favourably after institution of antifungal agents.
The diagnosis of IA-HLH is often missed unless there is a high ‘index of clinical suspicion’. It is important to arrive at an early diagnosis since the mortality and morbidity increase exponentially with a delay in diagnosis, which in turn delays institution of aggressive supportive care and specific treatment. Without specific treatment, HLH can be rapidly fatal. It has been recommended that activated macrophages/histiocytes should be targeted using IVIg, steroids, etoposide and activated T cells, by using steroids with or without ciclosporin A.18
Ideally, molecular diagnosis should be sought in all the cases, as primary HLH is now known to manifest also in adults. We established the diagnosis of IA-HLH in these 26 patients by using complete blood counts, biochemical parameters and demonstration of haemophagocytosis on bone marrow aspiration/biopsy studies. LDH was found to be raised in all our patients and was considered a useful tool in the diagnostic investigation. Serum ferritin was measured at screening and repeated if results were inconclusive or when a patient's clinical condition worsened. The biochemical parameters, including peak values of LDH, aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bilirubin, ferritin and lowest values of fibrinogen, showed a distinct difference between the fatal and non-fatal group, being worse in the former. However, Chen et al8 found significant differences only in peak values of AST, ALT and total bilirubin in the fatal cases as compared with non-fatal cases. Though demonstration of haemophagocytosis on bone marrow (BM) aspiration/biopsy is not mandatory for the diagnosis of HLH according to the diagnostic criteria, we found this to be a useful tool, in the absence of molecular studies. It is important to note that in four cases, serial BM aspiration/biopsy was needed to show >3% haemophagocytosis. This emphasises the importance of serial BM examination for the diagnosis of HLH.
We feel that immunomodulatory treatment should be started early to produce the best outcomes. This is of utmost importance in view of the resource limitations in developing countries, compounded by lack of universal availability of modern laboratory infrastructure. Clinical expertise is often the only weapon in the treating physician's armoury as the diagnosis has to be relentlessly pursued and may take from days to weeks before it is established. Steroids and IVIg are indicated for a broad range of HLH either alone or in combination with other treatments.19 ,20 Efficacy is variable, which is not surprising given the heterogeneity of this disorder. Many patients with EBV-associated HLH and those who do not respond to IVIg and steroids will require management with aggressive combination chemotherapy followed by haematopoietic stem cell transplant for better outcome.2
Published case reports show that the use of granulocyte-macrophage CSF (GM-CSF) is associated with worsening splenomegaly and proliferation of histiocytic cells in the bone marrow leading to worsening cytopenia.21 ,22 We also found that use of G-CSF in three patients led to deterioration in their clinical condition and all these patients died.
IA-HLH in adults is not uncommon as thought earlier and is associated with a wide spectrum of infections with variable outcome. In all patients presenting with a febrile syndrome with features of multisystem inflammation and varying degrees of unexplained pancytopenia, all efforts should be made to establish or rule out IA-HLH. Early recognition of IA-HLH in the tropical Asian populations should stimulate a search for common infections, with early institution of specific treatment and aggressive supportive care, which are crucial for reducing morbidity and mortality. From the clinical findings of the patients who died, it was also observed that the presence of comorbid conditions increases the risk of mortality as do the haematological and biochemical parameters described earlier.
Despite a high incidence of tropical infections in this part of the world, studies on IA-HLH are sparse and limited to case reports. The main reasons are lack of clinical suspicion and awareness and non-availability of genetic and other molecular studies in most developing countries. Diagnosis is usually delayed, which has a negative effect on the morbidity and mortality. Further studies will be needed to establish the significance of pretreatment values of various haematological and biochemical parameters to enable prediction of the high-risk subsets in this disorder.
Infection-associated haemophagocytic lymphohistiocytosis (IA-HLH) in adults is not uncommon as thought earlier and is associated with a wide spectrum of infections with variable outcome.
Though bone marrow haemophagocytosis is not a mandatory diagnostic criterion, we found it to be a useful tool for diagnosis of IA-HLH, in the absence of specialised laboratory facilities.
A clinical suspicion of IA-HLH in the tropical Asian populations should stimulate a search for common infections, with institution of specific antimicrobial agents and immunomodulatory treatment together with aggressive supportive care, which is crucial for reducing morbidity and mortality.
Growth factors such as granulocte colony stimulating factor should be avoided once HLH is suspected.
Current research questions
Prospective studies to establish the incidence of infection-associated haemophagocytic lymphohistiocytosis (IA-HLH) and document the pattern of infections in the Indian subcontinent.
Role of haematological, laboratory parameters and bone marrow examination in diagnosis and prognosis of IA-HLH
Need to evolve a treatment algorithm for IA-HLH
Rouphael NG, Talati NJ, Vaughan C, et al. Infections associated with haemophagocytic syndrome. Lancet Infect Dis 2007;7:814–22.
Filipovich AH. Hemophagocytic Lymphohystiocytosis (HLH) and related disorders. Hematology 2009;1:127–31.
Quesnel B, Catteau B, Aznar V, et al. Successful treatment of juvenile rheumatoid arthritis associated haemophagocytic syndrome by cyclosporin A with transient exacerbation by conventional-dose G-CSF. Br J Haematol 1997;97:508–10.
Wang S, Degar BA, Zieske A, et al. Hemophagocytosis exacerbated by GCSF/GM-CSF treatment in a patient with myelodysplasia. Am J Hematol 2004;77:391–96.
We thank O/O DGAFMS, DGMS (army) and Commandant Army Hospital, Research and Referral, New Delhi, India, for their support of this study. We also acknowledge Col (retd) DK Mishra, Col Jyoti Kotwal, Col T Chaterjee and Lt Col P Ganguli for their excellent laboratory support for this study. We thank our nursing and support staff for their dedicated work. Lastly, and most importantly, we are grateful to all our patients whose cooperation helped us to learn.
Contributors VN: study concept, design, management of patients, manuscript writing, revision and overall supervision. SD: data collection, patient management and manuscript writing. AS, SS: data collection and patient management. PS: manuscript writing, revision and data analysis. SR: manuscript revision and statistical analysis. SB: laboratory support.
Competing interests None.
Ethics approval The study was approved by the institutional ethics committee of Army Hospital (Research and Referral), New Delhi 110010.
Provenance and peer review Not commissioned; externally peer reviewed.
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.