SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Objective

To study how corticosteroid therapy is planned and described in reports of systemic disease trials and estimate the impact of the between-arm difference in corticosteroid dose on treatment effect.

Methods

We performed a systematic review of PubMed and Cochrane databases on all reports of randomized systemic disease trials with corticosteroids as a cointervention. Data were extracted on the trial characteristics and results, planning of corticosteroid use, and dose. Success rates were adjusted for corticosteroid use for studies with available data and a binary outcome. Because the exact impact of between-arm differences in corticosteroid dose on success rates is unknown, we tested different values for the impact of a difference of 1 unit (1 mg for daily dosage or 250 mg for cumulative dose at the end of the trial).

Results

A total of 139 trials were identified, including 79 investigating lupus and 30 investigating vasculitis. Planned management of corticosteroid use was specified in 101 reports (72.7%), with a fully described tapering scheme in 33 (23.7%). Corticosteroid consumption for each arm was given in 60 reports (43.2%), with a comparison of daily or cumulative dosage at the end of the trial in 32 (23.0%). An attempt to adjust for corticosteroid use was described in 2 (1.5%). With a value of 2.5% for the impact of a 1-unit difference in corticosteroid dose, adjustment yielded changes in success rate differences exceeding 10% in 11 (46%) of the 24 reports analyzed.

Conclusion

For systemic disease trials, use of corticosteroids as a cointervention is often inadequately planned and reported and could affect treatment effect.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Corticosteroids are considered the cornerstone of treatment of most systemic diseases such as systemic lupus erythematosus or vasculitides. Whether the combined prescription of another therapeutic agent (e.g., an immunosuppressant) can further improve patient outcome is evaluated by conducting clinical trials, whereby the treatments being compared are prescribed in addition to corticosteroids (1–4). Randomization aims to equally balance prior corticosteroid use in each arm at baseline, but corticosteroid dose is highly likely to be modified by physicians throughout the trial, depending on the course of the disease, the effect of the tested intervention, and adverse events (5). Therefore, because corticosteroids are a cointervention of major efficacy in this field, between-arm differences in corticosteroid consumption over the trial period may mask the superiority or, conversely, erroneously suggest the superiority of the experimental treatment.

Precise details of each active intervention should theoretically be described in reports of randomized clinical trials in order to interpret trial results. However, at present, no consensus or detailed recommendation exists on how to manage and take into account corticosteroid use when analyzing the results of trials of systemic diseases (6–8).

We conducted a systematic review of parallel-group randomized trials of systemic diseases in which corticosteroids were an essential part of the treatment. We studied how corticosteroid use was planned, managed, and monitored, and finally, how data on corticosteroid consumption were reported and taken into account when interpreting the study results. We also estimated the potential impact of between-arm differences in corticosteroid dose on study results.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Systematic literature search and study selection.

One of the authors (CP) searched the literature in the Medline database through PubMed using the keywords “randomized clinical trial” and systemic lupus erythematosus (“lupus,” “systemic lupus erythematosus,” “lupus nephritis,” or “neurolupus”), “vasculitis,” “polymyalgia rheumatica,” inflammatory myopathy (“myositis,” “inflammatory myopathy,” “polymyositis,” “dermatomyositis,” or “inclusion-body myositis”), “sarcoidosis,” or adult-onset Still's disease (“juvenile systemic arthritis,” “juvenile arthritis,” or “adult onset Still's disease”).

Eligible studies had to fulfill all of the following criteria: English or French articles 1) of human, parallel-group, randomized controlled clinical trials of systemic diseases, 2) comparing 2 agents or 1 agent versus placebo or control treatment, and 3) in combination with corticosteroids, for at least some of the patients, in both study arms. Reports of proceedings or those with an abstract only were excluded.

There was no restriction on publication date (Medline: 1966 to March 30, 2008). The Cochrane database was searched to ensure that all of the relevant studies had been identified. The reference lists of the retrieved articles were also examined for additional trials. Finally, for each selected article, we systematically searched and assessed online appendices and design articles.

Data collection.

One of the authors (CP) collected data on general characteristics of the trials, as well as data on corticosteroid therapy, using a standardized form. An independent reviewer (RS) screened a random sample of 10% of the selected articles to ensure agreement on the extracted data. We recorded whether patients had received corticosteroids prior to the study, and if so, at what dose; whether previous corticosteroid therapy was an inclusion criterion in the trial; whether the tapering or cessation of corticosteroids was chosen as the primary outcome (corticosteroid-sparing design) or as a secondary outcome; whether the management of corticosteroids throughout the study was planned as reported in the Methods section of the article, an appendix, or an earlier publication on trial design, and if yes, what the planned regimen for corticosteroids was (dose to remain stable, predefined tapering scheme, dose adjustment in case of disease flare or adverse event, or both); and whether corticosteroid initial, final, and cumulative doses were reported for each arm, compared and taken into account when interpreting the final study results.

The 2 investigators (CP, RS) collected and double checked quantitative data for all of the studies reporting a daily corticosteroid dosage at the end of the trial or a cumulative dose for each arm and having a binary outcome such as the rate of success or relapse.

Statistical analyses.

Descriptive data are described with numbers and percentages for qualitative variables and the mean or median (first and third quartiles) for continuous variables, as appropriate.

The extent to which a between-arm difference in daily corticosteroid dosage or cumulative dose received during the trial could affect the study results was estimated for all of the studies with a binary outcome and sufficient data reported (i.e., the mean or median cumulative or daily dosage at the end of the trial in each arm). The method used is detailed in Supplementary Appendix A (available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home). Briefly, we assumed that the therapeutic effect in each study arm depended on both the treatment under investigation and corticosteroid use. We assumed additive effects with no interaction, so we used a logistic statistical model with 2 covariates (treatment effect and corticosteroid dose). We formulated quantitative hypotheses regarding the association between corticosteroid consumption and efficacy. Corticosteroid use was assessed by the mean or median daily dosage at the end of the trial (with a prednisone equivalent dosage of 1 mg/day corresponding to 1 unit) or cumulative dose (with a prednisone equivalent dose of 250 mg corresponding to 1 unit). We therefore considered that a 1-unit between-arm difference in corticosteroid consumption was associated with a δ change for the success rate. The rate of success in the experimental arm could thus be adjusted for corticosteroid consumption, i.e., to a situation in which corticosteroid doses were the same in the 2 arms. We considered different plausible values for δ: 0.5%, 1%, 2.5%, 5%, 10%, and 15%. All of the statistical analyses involved the use of R software, version 2.6.1 (9).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

General characteristics of the selected studies.

The electronic search yielded reports of 139 studies (Figure 1 and Supplementary Appendix B, available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home). Their characteristics are given in Table 1. Briefly, the studies concerned mainly systemic lupus erythematosus (n = 79 [56.8%]) and vasculitis (n = 30 [21.6%]). A total of 109 trials (78.4%) included patients with active disease that was newly diagnosed, relapsing, or refractory, whereas 30 (21.6%) included patients with inactive disease but who were taking corticosteroids.

thumbnail image

Figure 1. Flow diagram of the selection process for articles.

Download figure to PowerPoint

Table 1. General characteristics of the 139 reports of systemic disease trials that involved the use of corticosteroids as a cointervention*
 Value
  • *

    Values are the number (percentage) of studies unless otherwise indicated. IQR = interquartile range.

Systemic diseases 
 Systemic lupus erythematosus79 (56.8)
 Polymyalgia rheumatica3 (2.2)
 Polymyalgia rheumatica and giant cell arteritis1 (0.7)
 Giant cell arteritis8 (5.8)
 Small- and medium-vessel vasculitides17 (12.2)
 Other vasculitides (Behçet's disease)4 (2.9)
 Inflammatory myopathies12 (8.6)
 Sarcoidosis9 (6.5)
 Adult-onset Still's disease6 (4.3)
No. of patients per study, median (IQR)38 (20–66)
Disease status 
 Active disease (newly diagnosed, relapsing, and/or refractory)109 (78.4)
 Inactive or moderately active disease30 (21.6)
Study design 
 Superiority135 (97.1)
 Noninferiority or equivalence4 (2.9)
Control group 
 Placebo (or no treatment) controlled123 (88.5)
 Active treatment controlled16 (11.5)
Blinding 
 No blinding67 (48.2)
 Single blind3 (2.2)
 Double blind69 (49.6)

Most of the trials were superiority trials (n = 135 [97.1%]) with placebo or no treatment as a control (n = 123 [88.5%]). One-half of the trials (n = 69 [49.6%]) were double blinded for the study treatments and 3 (2.2%) were single blinded. Tapering of corticosteroid therapy was the primary outcome measure in 9 trials (6.5%). Corticosteroid consumption was part of the secondary outcome measures in 25 trials (18.0%). A statistical analysis adjusted for corticosteroid use was described as planned in 2 other reports (1.5%), based on a Cox proportional hazards model adjusted for concomitant use of corticosteroids (10) or exploratory analysis of covariance adjusted for mean daily corticosteroid dosage (11).

Planned management for corticosteroids.

As shown in Table 2, information on planned management for corticosteroids was given for 101 studies (72.7%), but a fully detailed description of the predefined tapering scheme, including duration and dose for each interval, with or without adjustment according to therapeutic response, was given for only 33 studies (23.7%). Corticosteroid dose had to remain unchanged for another 31 trials (22.3%). Whether this scheduled scheme for corticosteroids was respected or not was reported for only 1 study, in which up to 20% of the physicians did not adhere to the planned algorithms for dose adjustment.

Table 2. Planning for corticosteroid therapy in 139 reports of randomized clinical studies of systemic diseases
 No. (%) of studies
  • *

    The tapering scheme was considered fully detailed when the duration and dose of each interval of corticosteroid tapering were specified. Other tapering schemes specified as timely scheduled but missing some or all of this information were considered partially detailed.

Planned starting dosage of corticosteroids 
 High (>0.5 mg/kg/day)50 (36.0)
 Intermediate (10 mg/day to 0.5 mg/kg/day)12 (8.6)
 Low (<10 mg/day)18 (13.0)
 Not fixed (high, medium, and/or low)26 (18.7)
 Not specified33 (23.7)
Planned scheme for corticosteroids 
 Yes101 (72.7)
  Stable dosage throughout the study period31 (22.3)
  Tapering adjusted to therapeutic response14 (10.1)
  Timely scheduled tapering (with or without adjustment of therapeutic response)56 (40.3)
   Fully detailed*33 (23.7)
   Partially detailed23 (16.6)
 Not specified38 (27.3)

Reporting on corticosteroid use prior to and during the trial.

For 102 studies, patients possibly received corticosteroids prior to inclusion. Among them, 73 articles (71.5%) described the duration and/or daily or cumulative dosages of prior corticosteroid therapy for each arm.

Information on corticosteroids received during the trial (i.e., duration, dose, or both) was given for only 60 (43.2%) of all 139 studies, including 38 (52.8%) of the blinded trials and 22 (32.8%) of the nonblinded trials (Table 3).

Table 3. Reporting of corticosteroid use in 139 reports of randomized clinical studies of systemic diseases
 All studies (n = 139)Blinded studies (n = 72)Nonblinded studies (n = 67)
  • *

    In 3 articles, results for comparison were given only in the Discussion section as statements, without any numerical data.

Studies reporting results for corticosteroid use during the trial, no. (%) of studies   
 At least 1 parameter reported for each arm60 (43.2)38 (52.8)22 (32.8)
  Mean or median daily starting dosage (baseline)32 (23.0)17 (23.6)15 (22.4)
  Mean or median daily dosage at the end of the trial24 (17.3)13 (18.1)11 (16.4)
   With mean/median daily starting dosage14 (10.1)7 (9.7)7 (10.4)
  Mean or median cumulative dose16 (11.5)9 (12.5)7 (10.4)
   With mean/median daily starting dosage7 (5.0)4 (5.6)3 (4.5)
  Mean or median duration of corticosteroid therapy6 (4.3)5 (6.9)1 (1.5)
  Proportion of patients weaned off of corticosteroids at the end of the trial6 (4.3)3 (4.2)3 (4.5)
  Proportion of patients under a target dose and/or achieving a target decrease of corticosteroids at the end of the trial8 (5.8)6 (8.3)2 (3.0)
 No parameter reported*79 (56.8)34 (47.2)45 (67.2)
Comparison between study arms for corticosteroid use during the trial (i.e., for 1 or more of the parameters listed above), no. (%) of studies   
 Yes53 (38.1)32 (44.4)21 (31.3)
 Not reported83 (59.7)39 (54.2)44 (65.7)
 Other*3 (2.2)1 (1.4)2 (3.0)
Comparison between study arms for daily corticosteroid dosage at the end of the trial or cumulative corticosteroid dose, no. (%) of studies   
 Yes32 (23.0)20 (27.7)12 (17.9)
  Between-arm difference exceeding 10%25 (18.0)16 (22.2)9 (13.4)
  Between-arm difference exceeding 30%9 (6.5)7 (9.7)2 (3.0)
 No107 (77.0)52 (72.2)55 (82.1)

As shown in Table 3, the daily corticosteroid dosage at the end of the trial was reported for 24 studies (17.3%) and the cumulative dose received during the trial was reported for 16 studies (11.5%), including 8 studies (5.8%) reporting both parameters.

Reporting of difference in corticosteroid use between arms.

Among the 60 studies reporting corticosteroid use for each arm during the trial, a statistical comparison was given for 53 (38.1%) and revealed a significant difference in 18 (34.0%) of them, with a relative between-arm difference for final daily or cumulative dosage exceeding 30% in 9 (17%) (Table 3). In the remaining studies for which between-arm differences for corticosteroid use were not statistically significant, the relative differences reached 31% for the final daily dosage in one trial and 28.7% for the mean cumulative dose in another.

Influence of corticosteroid use on study results.

As shown in Figure 1, we identified 25 studies with a binary outcome parameter and the required data on corticosteroid use to perform our statistical analysis. In one study, the reported mean corticosteroid daily dosage at the end of the trial was similar in each study arm. Six of the remaining 24 reports described the superiority and corticosteroid-sparing effect of one of the study agents.

Adjustment for corticosteroid dose altered the treatment effect by more than 10% in 4 of these 24 studies with a δ (i.e., the estimated difference in success rate with a 1-unit variation in corticosteroid dose) fixed at 1%, in 11 with a δ fixed at 2.5%, and in 14 with a δ fixed at 5%. In 17 of these 24 studies, the between-arm difference tended to increase after our statistical adjustment approach, whereas it was minimized in 5. In the 2 remaining studies, this adjustment even led to a reversal of the study results for success rate.

Figure 2 illustrates for some studies (12–17) how adjustment for the between-arm difference in corticosteroid dose may affect the study results in different ways when testing all prespecified values of δ (results for all 24 of the studies are shown in Supplementary Appendix C, available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home).

thumbnail image

Figure 2. Results of the statistical adjustment on corticosteroid (CS) consumption for the between-arm difference in rates of success in 6 studies (10–15) of systemic diseases with sufficient data to compute and a binary outcome (see the text for details on the selection of these studies). For each study, the open squares show the reported absolute differences for the rates of success between the experimental (Exp) and control (Ctrl) arms in each of the studies. The solid squares correspond to the absolute between-arm differences for the rates of success after adjustment with different ascending values (0.5%, 1%, 2.5%, 5%, 10%, and 15%) for the impact on the success rate of a difference of either 1 mg in the mean or median daily dosage of CS at the end of the trial or 250 mg in the mean or median cumulative dose of CS received during the trial (for details on the statistical analysis, see Supplementary Appendix B, available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home). Squares above the x axis correspond to a better success rate in the Exp arm, with ascending curves suggesting a shift toward a better than reported success rate with the experimental intervention. Conversely, squares under the x axis correspond to a better success rate in the Ctrl arm, with descending curves suggesting a shift toward a better than reported success rate of the treatment received by the control group. ANCA = antineutrophil cytoplasmic antibodies; IV = intravenous.

Download figure to PowerPoint

For the only 2 studies with a planned adjustment for corticosteroid consumption, the reported statistical adjustment did not alter the results of treatment effect. Our statistical model could be applied to one of these studies (study 23 in Supplementary Appendix C, available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home) (10) having a binary outcome parameter, and just confirmed the results reported by the authors (i.e., the superiority of the experimental treatment).

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Our systematic review of reports of clinical trials of systemic diseases showed that the planning of corticosteroid use as a cointervention was only partially detailed or not even reported in more than one-third of the trial reports. More importantly, corticosteroid doses received during the trial were reported for only one-half of the blinded studies and less than one-third of the nonblinded studies. Between-arm differences in corticosteroid dose exceeded 10% in more than one-half of the studies reporting these data, which potentially affected the study results for the primary outcome, as was suggested by our statistical analysis.

Corticosteroids are major active medications used in systemic diseases. However, as a cointervention, their potential effect on the results of parallel-group randomized trials has not been extensively evaluated (2, 3, 18, 19). Our study showed that the between-arm difference in corticosteroid dose may have a clinical impact and may lead to biased interpretation of the results.

Close monitoring and adequate reporting of corticosteroid use in trials is essential to allow adequate analysis of the combined study treatment effects in terms of efficacy and toxic effects. Randomization can limit the potential between-group imbalance for covariables, such as potential endogenous and some exogenous differences in corticosteroid metabolism, but only at baseline and not throughout the entire study duration. In addition, corticosteroid dose is particularly subject to change during patient followup. A similar regimen of corticosteroid consumption in each arm, in terms of both dose and duration, therefore represents the ideal setting to assure the comparability between arms in the final analysis. Therefore, if a predefined plan for corticosteroid management has been devised for a trial, the plan should be systematically reported. The proportion of patients who escape from disease control with the planned corticosteroid taper schedule could then represent a measurable outcome as part of the trial design. However, Baughman et al (20) showed that, even when trialists planned precise algorithms for corticosteroid dose changes during the trial, up to 20% of physicians do not strictly adhere to the plan. Indeed, management of corticosteroid dose in systemic diseases often relies on each physician's practice and may be influenced in nonblinded trials by knowledge of the treatments being compared. Therefore, blinding is critical to avoid or at least limit the risk of performance bias related to a differential corticosteroid use in either of the study arms.

Our results also suggested that adjustment for corticosteroid consumption in each group may modify the study results for the main comparison. Analysis of the results of randomized trials adjusted for covariates is usually limited to covariates measured at baseline (5). Adjustment for covariates measured after randomization is rare and is questionable, but may be warranted when these covariates have a strong correlation with the outcome (i.e., a prognostic value) or when there is a high degree of treatment imbalance between arms (21–23).

We acknowledge that our assumptions about the effect of a between-arm difference in corticosteroid dose on success rate remains questionable. Nevertheless, determining a single value that would be appropriate for all of the trials is impossible, because it may vary according to the disease, dose received during and still taken at the end of the trial (i.e., high versus low dose), or corticosteroid use before the trial. Therefore, we tested a wide range of plausible values for this estimated impact of corticosteroids. In addition, this value may vary according to potential drug interactions and/or individual pharmacogenetic profiles that are only partially or not known yet, but may deserve additional adjustments. However, since we did not have full access to raw data, we did not consider potential interactions between the steroid effect and the nonsteroid effect. Finally, adjustment on corticosteroid use was performed on global study results and not on individual data. Therefore, we could not test the treatments being compared after adjustment for corticosteroid use. Notably, the estimation of the effect of the between-arm difference in corticosteroid dose on success rate for individual data may yield different and possibly even more demonstrative results.

Our study showed that planning for and reporting corticosteroid use in clinical trials of systemic diseases is often not adequately addressed, which raises concerns about the interpretation of the final results of these trials. No specific consensus or recommendation exists for designing, conducting, and reporting the results of trials in which corticosteroids are used as a cointervention. Because such trials are numerous and increasing in frequency, several methodologic points, listed in Table 4, should be systematically considered to limit this potential cointervention bias.

Table 4. Practical recommendations for planning and reporting results from randomized controlled trials with corticosteroids used as a cointervention
Points to considerRecommendations for planningRecommendations for reporting
Inclusion criteriaDefine whether corticosteroid use prior to study enrollment is allowed or not. If yes, define any limitations regarding the dose or duration of prior corticosteroid use.Report eligibility criteria concerning prior corticosteroid use in the Materials and Methods section.
RandomizationConsider stratified randomization on prior corticosteroid use.Report whether stratification is planned for prior corticosteroid use.
BlindingConsider blinding as essential to homogenize corticosteroid management and use in both arms.Report blinding status for all of the subjects involved in the study.
Primary outcomeConsider a corticosteroid-sparing design, in which efficacy and corticosteroid consumption are both part of the primary outcome. 
Corticosteroid use during the trialEstablish a standardized protocol for management of corticosteroid use throughout the study: initial dose, dose adjustment.Report this protocol in the Materials and Methods section or as an appendix.
Monitoring of corticosteroid use during the trialRecord corticosteroid use and dose for each patient throughout the study.Report corticosteroid dose by arm at entry and at the end of the study, as well as a time-cumulative parameter of the dose.
  Report the number of patients who deviated from the planned regimen and for what reasons.
AnalysisPlan to investigate the potential effect for corticosteroid consumption.Report adjustment for corticosteroid consumption in the Statistical Analysis section.
  Discuss the potential effect of corticosteroid consumption on the study results.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Pagnoux had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Pagnoux, Giraudeau, Guillevin, Ravaud.

Acquisition of data. Pagnoux, Seror.

Analysis and interpretation of data. Pagnoux, Dechartres, Giraudeau, Seror, Ravaud.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. REFERENCES
  9. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
ACR_20139_sm_appenda.doc32KSUPPLEMENTARY APPENDIX A: DETAILED STATISTICAL ANALYSIS FOR ADJUSTMENT ON CORTICOSTEROID CONSUMPTION
ACR_20139_sm_appendb.doc61KSUPPLEMENTARY APPENDIX B: COMPLETE REFERENCE LIST OF THE SELECTED RANDOMIZED TRIALS, FOR EACH DISEASE SEPARATELY LUPUS
ACR_20139_sm_appendc.doc14689KSUPPLEMENTARY APPENDIX C: RESULTS OF 24 STUDIES

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.