Article Text

PDF

Extended report
Definition and initial validation of a Lupus Low Disease Activity State (LLDAS)
  1. Kate Franklyn1,
  2. Chak Sing Lau2,
  3. Sandra V Navarra3,
  4. Worawit Louthrenoo4,
  5. Aisha Lateef5,
  6. Laniyati Hamijoyo6,
  7. C Singgih Wahono7,
  8. Shun Le Chen8,
  9. Ou Jin9,
  10. Susan Morton10,
  11. Alberta Hoi1,
  12. Molla Huq11,
  13. Mandana Nikpour11,
  14. Eric F Morand1
  15. for the Asia-Pacific Lupus Collaboration
  1. 1School of Clinical Sciences at Monash Health, Monash University, Melbourne, Victoria, Australia
  2. 2University of Hong Kong, Hong Kong, Hong Kong
  3. 3University of Santo Tomas, Manila, Philippines
  4. 4Chang Mai University, Chang Mai, Thailand
  5. 5National University Hospital, Singapore, Singapore
  6. 6Padjadjaran University, Bandung, Indonesia
  7. 7Brawijaya University, Malang, Indonesia
  8. 8Shanghai Jiaotong University, Shanghai, China
  9. 9Division of Rheumatology, Affiliated 3rd Hospital of Sun Yat-san University, Guangzhou, China
  10. 10Monash Health, Monash University, Melbourne, Victoria, Australia
  11. 11Department of Medicine, The University of Melbourne, Melbourne, Victoria, Australia
  1. Correspondence to Professor Eric F Morand, School of Clinical Sciences at Monash Health, Monash University, Level 5, Block E, Monash Medical Centre, 246 Clayton Road, Clayton, Melbourne, VIC 3168, Australia; eric.morand{at}monash.edu

Abstract

Aims Treating to low disease activity is routine in rheumatoid arthritis, but no comparable goal has been defined for systemic lupus erythematosus (SLE). We sought to define and validate a Lupus Low Disease Activity State (LLDAS).

Methods A consensus definition of LLDAS was generated using Delphi and nominal group techniques. Criterion validity was determined by measuring the ability of LLDAS attainment, in a single-centre SLE cohort, to predict non-accrual of irreversible organ damage, measured using the Systemic Lupus International Collaborating Clinics Damage Index (SDI).

Results Consensus methodology led to the following definition of LLDAS: (1) SLE Disease Activity Index (SLEDAI)-2K ≤4, with no activity in major organ systems (renal, central nervous system (CNS), cardiopulmonary, vasculitis, fever) and no haemolytic anaemia or gastrointestinal activity; (2) no new lupus disease activity compared with the previous assessment; (3) a Safety of Estrogens in Lupus Erythematosus National Assessment (SELENA)-SLEDAI physician global assessment (scale 0–3) ≤1; (4) a current prednisolone (or equivalent) dose ≤7.5 mg daily; and (5) well tolerated standard maintenance doses of immunosuppressive drugs and approved biological agents. Achievement of LLDAS was determined in 191 patients followed for a mean of 3.9 years. Patients who spent greater than 50% of their observed time in LLDAS had significantly reduced organ damage accrual compared with patients who spent less than 50% of their time in LLDAS (p=0.0007) and were significantly less likely to have an increase in SDI of ≥1 (relative risk 0.47, 95% CI 0.28 to 0.79, p=0.005).

Conclusions A definition of LLDAS has been generated, and preliminary validation demonstrates its attainment to be associated with improved outcomes in SLE.

  • Systemic Lupus Erythematosus
  • Disease Activity
  • Outcomes research

Statistics from Altmetric.com

Introduction

For patients afflicted by systemic lupus erythematosus (SLE), mortality remains unacceptably high.1 Organ damage, morbidity and mortality in SLE are the result of disease activity and/or the consequences of therapy,2–4 but as well as a paucity of specific therapies, a lack of metrics to guide the management of SLE continues to hinder clinicians’ ability to influence outcomes. Although a number of SLE disease indices have been validated, they possess widely acknowledged flaws,5 ,6 arising in part from the inherent heterogeneity of SLE. For example, the SLE Disease Activity Index (SLEDAI)5 lacks the ability to measure severity of disease activity within an organ system, and indeed omits entire organs such as the gastrointestinal tract. The British Isles Lupus Assessment Group (BILAG) 2004 is a more comprehensive instrument, but a high level of detail renders it cumbersome for use in everyday practice.6

Treatment approaches focused on achieving predetermined states have been demonstrated to improve patient outcomes across a wide spectrum of medical conditions.7–9 Strategy trials in rheumatoid arthritis (RA) have revealed the utility of a low disease activity state (LDAS) definition, now widely adopted in clinical practice as a foundation of treat-to-target approaches.10 Importantly, the utility of a low disease activity target in RA was demonstrated prior to the introduction of biological therapies, and in diseases such as hypertension, the adoption of a target, rather than the use of novel therapies, is the cornerstone of treat-to-target approaches.7–9 Although recent years have seen the introduction of SLE responder states in drug trials,11–13 these record changes from baseline rather than a target state. Reported remission definitions in SLE are varied, and none is universally accepted, but the most stringent are achieved by less than 1 in 50 patients and are impractical for use in clinical trials or daily practice.14 Therefore, there is a need for the definition of a low disease activity state in SLE, which in turn enables treat-to-target strategies to be tested in this disease.15 ,16 Here, we have generated a consensus definition of a lupus LDAS (LLDAS). Preliminary validation of this definition of LLDAS suggests that it is associated with prediction of improved patient outcomes.

Methods

Definition of LLDAS

The LLDAS definition was generated by a panel of lupus experts from Hong Kong, China, Philippines, Thailand, Singapore, Indonesia and Australia (figure 1). The panel first agreed on a conceptual definition of LLDAS, as ‘a state, which if sustained, is associated with a low likelihood of adverse outcome, considering disease activity and medication safety’. Delphi and nominal consensus techniques, which support face and content validity, were then used to generate a definition for thresholds at or below which this would be considered achieved. The panel was invited to contribute items for potential inclusion in a definition of LLDAS based on their clinical experience and review of the literature. Fifty-six items for potential inclusion in the definition of LLDAS were contributed, in two domains: (1) disease activity and (2) medication safety. Using the Delphi method, these items were each scored independently by panel members on a 5-point scale (1=Strongly Disagree, 2=Disagree, 3=Unsure, 4=Agree, 5=Strongly Agree) using an online survey tool. Items with a mean score of >3 were retained.

Figure 1

Lupus Low Disease Activity State (LLDAS) consensus definition methodology. This figure summarises the process used to reach a consensus definition of LLDAS. After initially agreeing on a conceptual definition, all contributors independently generated items for potential inclusion in the LLDAS definition. These items underwent two rounds of Delphi and a nominal group discussion, resulting in a final list of five non-redundant items with very high agreement.

The second phase of the definition process involved 11 experts who comprise the authors of this study. The nominal group technique, which provides an orderly procedure for obtaining qualitative information from a group of experts, was followed by a second round of Delphi, where the remaining items were again scored out of 5. Items with a mean score greater than 4, indicating high levels of agreement, were retained. This process resulted in the final 5-item composite LLDAS definition (table 1).

Table 1

LLDAS definition

LLDAS validation

Criterion validity was sought by testing the association of LLDAS with outcomes. Given the absence of analogous low disease activity scores, damage accrual, measured using the Systemic Lupus International Collaborating Clinics Damage Index (SDI),17 was used as the primary end point for validation. We also evaluated the association of LLDAS with the adjusted mean SLEDAI-2K18 and future flares19 as secondary end points.

The data set comprised a prospective longitudinal observational cohort of patients with SLE established in 2007 in Melbourne, Australia, as described.20 ,21 Patients assessed between 2007 and 2012 who met the American College of Rheumatology (ACR) 1997 classification criteria for lupus22 were eligible for inclusion if they were over 18 years of age, were competent to provide informed consent, had been seen on at least two occasions, and had at least one SDI completed. Information collected at baseline included demographic data such as age, date of diagnosis by a qualified rheumatologist, year of first symptom, ACR classification criteria, ethnicity, antinuclear antibody titre, extractable nuclear antigen antibodies, anti-dsDNA and antiphospholipid antibody titres, and complement (C3 and C4) levels.

Visit frequency was determined by clinical need. At every visit including baseline, the SLEDAI-2K score,18 physician global assessment (PGA) (range 0–3), Safety of Estrogens in Lupus Erythematosus National Assessment (SELENA)--SLEDAI flare index19 and all SLE medication data were collected. Further laboratory data including a full blood count, urea and electrolytes, liver function tests, complement levels, anti-dsDNA, urine protein:creatine ratio and urinary sediment were also recorded. Documentation of the SDI was completed at baseline and annually.

Data analysis

Retrospective analysis of the prospectively collected data was performed. Achievement of LLDAS was determined at each visit using the aforementioned criteria. If a patient was in LLDAS on a given visit, for the purposes of analysis they were considered in LLDAS until the day prior to their next review. Where SLEDAI-2K data were incomplete due to missing laboratory investigations, the score for that item was assumed to be zero. Given the absence of information in the database on doses of immunosuppressant prior to the first visit, we assigned patients at the first visit as meeting criteria for LLDAS if they met all other criteria and the dose of the medication was within an acceptable dosage range. Where annual SDI data was missing, the last value was carried forward, provided patients had at least two SDI results recorded prior to the missing value.

As patients had varying lengths of follow-up in this cohort study, we evaluated the effect of LLDAS duration on outcomes in terms of the cumulative percentage of follow-up time spent in LLDAS. We initially defined the cut-off for ‘short’ and ‘long’ durations spent in LLDAS using the median (50%) of the observed period, and performed sensitivity analyses on other durations.

Univariable analysis was first performed to assess whether LLDAS was associated with reduction in damage accrual. Student's t tests, Mann–Whitney and χ2 tests were used where appropriate based on the distribution of data. After ensuring proportionality of hazard, ‘repeated failures’ proportional hazards models were used to determine the time-dependent relationship between LLDAS and flares at subsequent visits. In this way, we were able to determine whether LLDAS presence at each and every visit in each and every patient was predictive of outcomes (flare and damage accrual) at each and every subsequent visit. Flares were divided into severe flares, mild-to-moderate flares and a composite of any flare, according to the SELENA-SLEDAI flare index definitions.19 We used linear regression models to determine the relationship between achievement of LLDAS and an increase in SDI of at least one point, which was deemed clinically significant.23 A log binomial regression model was used to determine the relationship between proportion of time in LLDAS and change in SDI during follow-up. Finally, a sensitivity analysis was performed where we changed the prednisolone dose in the definition of LLDAS from ≤7.5 mg daily to ≤5 mg daily and determined the relationship of this definition of LLDAS with subsequent flare and damage accrual as above.

Results

A consensus definition of LLDAS

Fifty-six items were initially generated (see online supplementary table S1). The highest scored items in the Delphi survey related to disease activity, and can be summarised as a requirement for low and stable disease activity. The highest ranked item, with unanimous agreement (a score of 5), was disease activity measured using the SLEDAI-2K24 <4, excluding SLEDAI-2K activity in major organ systems (renal, central nervous system (CNS), cardiopulmonary, vasculitis, fever), and excluding haemolytic anaemia or gastrointestinal activity (these being absent from the SLEDAI-2K instrument) (table 1). The next highest ranked item was a SELENA-SLEDAI PGA (scored 0–3)19 of <1 (table 1), followed by the exclusion of any current disease activity that was new compared with the previous assessment. The next highest ranked items pertained to medication use. A requirement for current prednisolone (or equivalent) dosing of <7.5 mg daily, and for immunosuppressive drugs and/or approved biological agents (excluding investigational drugs), if used, to be at standard maintenance doses and without intolerance, scored equally (table 1). As lower ranked items had a mean rating score <4, or were largely redundant after inclusion of these five items, a five-part definition of LLDAS was agreed upon where all components are required simultaneously in order for LLDAS to be achieved. Duration of any variable was removed during the nominal group discussion, on the basis that insufficient data exist in the literature to determine what duration was most appropriate to incorporate, and to allow for time to be the subject of future analysis. In each case, the cut points are intended as thresholds, such that status at or below these levels was considered consistent with LLDAS, similar to the approach taken for the ACR-European League Against Rheumatism (EULAR) definition of remission in RA.25

Initial validation of LLDAS

We first determined the frequency of observation of LLDAS, using the definition above, in the 191 patients who met inclusion criteria (table 2) in whom 3352 visits were available for analysis. The average duration of disease at recruitment was 7.3 years, with an age at diagnosis of 33.3 years. All patients were antinuclear antibody positive, with 67% demonstrating anti-dsDNA positivity at recruitment. Our population comprised patients predominantly of European (58.1%) and Asian (39.5%) descent. The mean SLEDAI-2K at recruitment was 5.1, with nearly 80% of patients having musculoskeletal manifestations and just over 50% having evidence of lupus nephritis prior to or during the study. Patients were followed on average for 3.9 years (minimum 0.06 years, maximum 6.9 years) (table 2).

Table 2

Patient characteristics in LLDAS validation study (n=191)

As previously reported,20 ,21 patients in this cohort had evidence of active disease and of serological activity, with 97% receiving immunosuppressant drugs during follow-up and a mean daily dose of prednisolone of 7.07 mg (table 2). We analysed 3352 visits, representing a mean (SD) of 17.5 (11.7) visits per patient over 3.9 (2.0) years of follow-up (table 3).

Table 3

Disease activity and LLDAS attainment during follow-up

LLDAS was achieved on at least one occasion by 88.5% of patients. Online supplementary table S2 shows the number (%) of visits fulfilling each of the five LLDAS criteria. No two LLDAS criteria were completely concordant in this data set. Overall, among 3352 visits, 51(1.52%) visits fulfilled no criteria, 3301 (98.48%) visits fulfilled one criterion, 3078 (91.83%) visits fulfilled two criteria, 2651 (79.09%) visits fulfilled three criteria, 1983 (59.16%) visits fulfilled four criteria and 1132 (33.37%) visits fulfilled all five LLDAS criteria.

The median cumulative duration of LLDAS among all patients was 1.6 years, representing 41% of the total period of observation, and the mean duration of individual episodes of LLDAS was 106 days (table 3). Analysis of various cumulative durations of LLDAS revealed that 50% of total observed time spent in LLDAS was the optimal cut point for differentiating between groups in terms of outcome (moderate, severe and total flares, and damage accrual) (see online supplementary table S3). Patients who were in LLDAS ≥50% of the observed time were significantly older at diagnosis and more likely to be male, had significantly lower disease activity at recruitment, and trended towards having shorter disease duration at enrolment, but there was no statistically significant difference in ethnicity (table 4). As expected, patients who were in LLDAS ≥50% of the time had significantly lower prednisolone dose during follow-up, but also used significantly fewer immunosuppressant drugs (table 4). Disease activity measured using adjusted mean SLEDAI-2K, which accurately reflects disease activity over time when intervals between assessments are variable,18 was significantly lower in patients who were in LLDAS ≥50% of the time (table 4). Multiple specific manifestations of SLE were significantly less frequent in patients who were in LLDAS ≥50% of the time, including vasculitis, neurological, renal, cardiopulmonary and mucocutaneous manifestations; no significant difference in musculoskeletal or haematological manifestations was observed (table 4). Univariable analysis also revealed a significant reduction in future flares among patients who spent greater than 50% of their time in LLDAS, including mild/moderate, severe and all flares (p<0.0001) (table 4). The effect of LLDAS attainment on subsequent flare was further analysed using Cox proportional hazards regression, which revealed statistically significantly lower hazard of any future flare (HR 0.63, CI (0.52 to 0.77), p<0.0001) or severe flare (HR 0.18 (CI 0.10 to 0.30), p<0.0001) in those who spent ≥50% of follow-up in LLDAS (table 5). A significant reduction in all flares was also observed in multivariable analysis of the effect of total duration of LLDAS (table 6).

Table 4

Univariable analysis of characteristics associated with LLDAS among 191 patients with SLE

Table 5

The association of LLDAS with subsequent flare

Table 6

The association of LLDAS duration with flares, need for immunosuppressive treatment and damage accrual, during follow-up

Lastly, we analysed the effect of LLDAS attainment on damage accrual. In the cohort as a whole, a mean (SD) increase in SDI of 0.54 (1.16) resulting in a mean (SD) final SDI of 1.38 (1.79) was observed. Patients who spent ≥50% of observed time in LLDAS demonstrated a significantly lower increase in SDI, with a mean increase in SDI of 0.57 units less than that observed in patients with less time in LLDAS (table 4). Consistent with this finding, the relative risk (RR) for an increase in SDI over the study duration of >1 was 0.47 (95% CI 0.28 to 0.79, p=0.005) (table 5), meaning that patients who spent over 50% of the observed time in LLDAS had a twofold reduction in the risk of accruing a clinically significant change in SDI. When we substituted a prednisolone dose of ≤5 mg/day into the LLDAS definition, the RR for an increase in SDI >1 was 0.38 (95% CI 0.21 to 0.70, p=0.002). No deaths were observed during follow-up to permit evaluation of the relationship between LLDAS attainment and mortality.

Discussion

The use of treat-to-target approaches based on clinically achievable goals in numerous areas of medicine, including hypertension, diabetes mellitus and RA, has resulted in significant improvements in patient outcomes.7–10 As a consequence of the lack of defined treatment outcome states, establishing clear treatment guidelines and ‘treat-to-target’ approaches in SLE remains elusive. Determination of a measure of low disease activity for SLE has been recently identified as a major research priority.15 ,16 In the absence of a defined treatment goal, clinicians generally form an overall clinical impression based on balancing clinical and laboratory variables, treatment effects and the patient's impression of disease activity, to determine if a given state of disease is acceptable. In recent years, the need to measure treatment effects in clinical trials has led to innovations such the SLE responder index and the BILAG-based Composite Lupus Assessment instrument.11–13 However, these assess change in clinical status from a given starting point, rather than attainment of a target clinical state. A stringent definition of remission in SLE, proposed as sustained complete absence of disease activity without maintenance immunosuppressive treatment, was achieved by less than 2% of patients studied,14 rendering this definition impractical for clinical or research purposes. We therefore recently proposed that attainment of low disease activity could represent an alternative treatment target in SLE.15 ,16

In defining LLDAS, the objective was to develop an operational definition of our initial concept definition. The strengths and shortcomings of a number of contemporary disease activity instruments were considered. Although the BILAG6 was considered, its comprehensive format was considered unfeasible for routine clinical use. In contrast, the SLEDAI-2K was considered by the contributors to have high potential utility in clinical and research settings. The consensus definition methodology unanimously yielded <4 as the preferred SLEDAI-2K cut-off, with the condition that disease activity in major organ systems measured in the SLEDAI-2K, and in important systems not measured in the SLEDAI-2K such as gastrointestinal involvement and haemolytic anaemia, were excluded. The contributors’ view that new disease activity even in a permitted organ system was not consistent with the desired clinical state is reflected in excluding from LLDAS any disease activity not present at the previous assessment.

The use of a composite measure, which ensures against the limitations of any one instrument, is supported by Outcome Measures in Rheumatology (OMERACT) recommendations. The virtual impossibility of any instrument containing every possible manifestation of SLE led those who derived the SLE responder index and BILAG-based Composite Lupus Assessment to include a PGA in these composite measures.11–13 Our methodology yielded a high level of agreement for inclusion of a PGA (0–3) of <1 in the definition of LLDAS.

The contributors considered that inclusion of medication status was important to the definition of LLDAS, on the basis that low disease activity in the setting of high-dose glucocorticoids, which are known to contribute to morbidity in SLE,26 or of adverse effects from immunosuppressant drugs, is not a desirable clinical state. Our process yielded a definition of LLDAS that included glucocorticoid dose and immunosuppressant medication tolerability alongside disease activity domains. During the Delphi process, there was much debate among contributors regarding the cut point for prednisolone dose for this LLDAS item. Studies by Petri et al suggested a dose of 6 mg/day or less was associated with reduced damage accrual, but the dose range to which this was compared included doses up to twice as high27; a cut-off of <7.5 mg/day of prednisolone or equivalent glucocorticoid was most highly ranked by the contributors. Acknowledgement of the diversity of immunosuppressant dugs used in SLE and of their drug-specific adverse events resulted in a high ranking for an item permitting the use of immunosuppressive drugs and approved biologicals in standard doses, not resulting in significant intolerance, in the definition of LLDAS.

The results of our analyses are highly encouraging for the validity and potential usefulness of the current definition of LLDAS. First, no two criteria were completely concordant in the cohort studied. Second, LLDAS attainment was readily demonstrated in a proportion of patients, suggesting the measure has utility. Third, compared with patients who spent a minority of their observed time in LLDAS, patients spending the majority of their observed time in LLDAS had significantly lower frequency of SLE flare and accrual of organ damage, pointing to criterion validity. The sample size did not permit analysis of the ‘minimum’ amount of time necessary to spend in LLDAS to improve outcome, the effect of consecutive versus non-consecutive visits in LLDAS, or the optimum level of low disease activity to impact on damage accrual, and such analysis should be undertaken using a larger data set. Similarly, the sample size did not permit analysis of the optimum glucocorticoid dose associated with reduction in damage accrual. To explore this potentially controversial definition criterion, we substituted a prednisolone dose of ≤5 mg/day into the LLDAS definition for outcomes analysis and found that the RR for an increase in SDI >1 was 0.38 (95% CI 0.21 to 0.70, p=0.002) compared with a RR of 0.47 from the consensus definition of LLDAS (≤7.5 mg/day). This suggests that a more stringent threshold for prednisolone dose in a definition of LLDAS could be associated with further protection from organ damage accrual, an issue which requires resolution in a larger prospective study. Indeed, the authors and colleagues have commenced a prospective multicentre study to evaluate the association between achieving LLDAS and outcome, which will permit validation of the current LLDAS definition, and may permit further refinement of the definition.

The current study is limited by the retrospective nature of the validation analyses, undertaken in a cohort of patients followed at a single centre. However, all parameters included in the consensus LLDAS definition were included in the data set analysed, which had been collected prospectively. Further, our assumption regarding the duration of LLDAS from one visit to the next may overestimate the duration of time spent in LLDAS, but if that were the case, it would weaken the potential for these data to reveal a protective association of LLDAS with damage outcomes. In addition, one potentially important omission from the consensus definition of LLDAS is the lack of a patient reported outcome (PRO) or quality of life measure. However, given the lack of concordance with disease activity in several studies evaluating PROs in lupus,28 the investigators considered that there were insufficient data to support inclusion of a PRO in the definition of LLDAS at this time. As the definition of LLDAS was reached without reference to the characteristics of the cohort to be analysed, the high number of patients on immunosuppressants and the absence of patients on biologicals mean that evaluation of this component of item 5 of the LLDAS definition is incomplete. Prospective studies in which each domain of this definition of LLDAS, in particular glucocorticoid dose and the effect of specific immunosuppressant drugs, are subjected to sensitivity analysis could in the future yield a more sensitive and specific definition of LLDAS.

In conclusion, responding to the need for definition of treatment targets in SLE, we have generated a consensus definition of LLDAS, the attainment and cumulative duration of which is associated with reduction in flares and accrual of damage-related morbidity, supporting the validity of LLDAS as a predictor of outcome in SLE. The definition of LLDAS could have far-reaching ramifications for the development of treat-to-target paradigms in SLE, and a broad range of clinical and research applications.

Acknowledgments

The authors thank Dr Diane Apostopoulos for her assistance with data analysis.

References

View Abstract

Review history and Supplementary material

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

  • Lay summary

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Handling editor Tore K Kvien

  • EFM and MN contributed equally.

  • Correction notice This article has been corrected since it was published Online First. Reference 10 has been corrected.

  • Contributors All authors as listed below, KF, CSL, SVN, WL, AL, LH, CSW, SLC, OJ, SM, AH, MH, MN, EFM, were involved in all the following aspects of the research presented: Substantial contributions to the conception or design of the work, or the acquisition, analysis or interpretation of data. Drafting the work or revising it critically for important intellectual content. Final approval of the version published. Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

  • Funding KF has received a scholarship from Arthritis Australia. MN is a recipient of a David Bickart Clinician Research Fellowship from the University of Melbourne Faculty of Medicine, Dentistry and Health Sciences and holds an NHMRC Research Fellowship (APP1071735). EFM's research is supported by Arthritis and Osteoporosis Victoria. The AsiaPacific Lupus Collaboration is supported by grants from GlaxoSmithKline, UK and UCB Biopharma SPRL, Belgium.

  • Competing interests None declared.

  • 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.