Article Text


Extended report
Long-term safety of tofacitinib for the treatment of rheumatoid arthritis up to 8.5 years: integrated analysis of data from the global clinical trials
  1. Stanley B Cohen1,
  2. Yoshiya Tanaka2,
  3. Xavier Mariette3,
  4. Jeffrey R Curtis4,
  5. Eun Bong Lee5,
  6. Peter Nash6,
  7. Kevin L Winthrop7,
  8. Christina Charles-Schoeman8,
  9. Krishan Thirunavukkarasu9,
  10. Ryan DeMasi10,
  11. Jamie Geier10,
  12. Kenneth Kwok10,
  13. Lisy Wang11,
  14. Richard Riese11,
  15. Jürgen Wollenhaupt12
  1. 1Metroplex Clinical Research Center, Dallas, Texas, USA
  2. 2University of Occupational and Environmental Health, Kitakyushu, Japan
  3. 3Paris-Sud University, Le Kremlin Bicêtre, France
  4. 4University of Alabama at Birmingham, Birmingham, Alabama, USA
  5. 5Seoul National University, Seoul, Republic of Korea
  6. 6University of Queensland, Queensland, Australia
  7. 7Oregon Health and Science University, Portland, Oregon, USA
  8. 8University of California, Los Angeles, California, USA
  9. 9Pfizer Australia, Sydney, Australia
  10. 10Pfizer Inc, New York, New York, USA
  11. 11Pfizer Inc, Groton, Connecticut, USA
  12. 12University of Hamburg, Hamburg, Germany
  1. Correspondence to Dr Ryan DeMasi; Pfizer Inc, 235 E 42nd St, New York, NY 10017, USA; Ryan.DeMasi{at}


Objectives Tofacitinib is an oral Janus kinase inhibitor for the treatment of rheumatoid arthritis (RA). We report an integrated safety summary of tofacitinib from two phase I, nine phase II, six phase III and two long-term extension studies in adult patients with active RA.

Methods Data were pooled for all tofacitinib-treated patients (data cut-off: 31 March 2015). Incidence rates (IRs; patients with event/100 patient-years) and 95% CIs are reported for adverse events (AEs) of interest.

Results 6194 patients received tofacitinib for a total 19 406 patient-years' exposure; median exposure was 3.4 patient-years. IR (95% CI) for serious AEs was 9.4 (9.0 to 9.9); IR for serious infections was 2.7 (2.5 to 3.0). IR for (all) herpes zoster was 3.9 (3.6 to 4.2); IR for disseminated or multidermatomal herpes zoster was 0.3 (0.2 to 0.4). IR for opportunistic infections (excluding tuberculosis) was 0.3 (0.2 to 0.4) and was 0.2 (0.1 to 0.3) for tuberculosis. IR for malignancies (excluding non-melanoma skin cancer (NMSC)) was 0.9 (0.8 to 1.0); NMSC IR was 0.6 (0.5 to 0.7). IR for gastrointestinal perforations was 0.1 (0.1 to 0.2). Analysis of IR for serious infections, herpes zoster and malignancies by 6-month intervals did not reveal any notable increase in IR with longer-duration tofacitinib exposure.

Conclusion This analysis of tofacitinib exposure up to 8.5 years allowed estimation of safety events with improved precision versus previous tofacitinib reports. AEs were generally stable over time; no new safety signals were observed compared with previous tofacitinib reports.

Statistics from


Tofacitinib is an oral Janus kinase inhibitor for the treatment of rheumatoid arthritis (RA). Tofacitinib has demonstrated efficacy and manageable safety in patients with active RA in phase I–III trials and long-term extension (LTE) studies.1–18 Results from these studies and adverse event (AE) profiles of other disease-modifying antirheumatic drugs (DMARDs) informed the selection of safety events of special interest: serious infection events (SIEs), opportunistic infections (OI; including tuberculosis (TB)), herpes zoster (HZ), malignancies, cardiovascular events and gastrointestinal (GI) perforations.

We report for the first time integrated data based on cumulative tofacitinib exposure throughout the RA development programme. This analysis extends previous safety reports for tofacitinib which focused on specific AEs and includes up to 8.5 years of tofacitinib exposure, allowing estimation of rates for safety events of interest with improved precision versus previous reports, and novel methods to examine dose-related AEs.

Patients and methods


Data were pooled from patients with RA treated with tofacitinib in phases I–III and LTE studies (see online supplementary table S1). All studies were completed by 31 March 2015 except LTE study NCT00413699. LTE data collection and analyses are ongoing (database not locked; some values may change in final locked database).

Patients (aged ≥18 years) had an active RA diagnosis based on the American College of Rheumatology 1987 revised criteria19 and active disease at screening and baseline.1–18 Key exclusion criteria included untreated infection with Mycobacterium tuberculosis or clinically significant infection, and history of malignancy except adequately treated squamous cell or basal cell skin cancer or cervical carcinoma in situ.1–18

Studies were conducted in compliance with the Declaration of Helsinki, International Council for Harmonisation Guidelines for Good Clinical Practice and local country regulations. The study protocol was approved by the Institutional Review Board or Independent Ethics Committee at each centre. Patients provided written informed consent.


Patients received tofacitinib 1, 3, 5, 10, 15 or 30 mg twice daily or 20 mg once daily, as monotherapy or with background DMARDs (see online supplementary table S1). Upon entering LTE studies, patients from phase II and III studies received tofacitinib 5 and 10 mg twice daily, respectively, regardless of index study treatment. Within the LTE, tofacitinib could be increased to 10 mg twice daily for inadequate response or reduced to 5 mg twice daily for safety reasons.

Because patients could change doses between index and LTE studies and within LTE, dose was categorised using two methods. The primary analysis used the average dosing algorithm in which patients were assigned to average tofacitinib 5 or 10 mg twice daily if the average daily dose at end of enrolment up to the cut-off date was <15 or ≥15 mg, respectively. The constant dosing algorithm was used in a sensitivity analysis. Only patients exposed to a constant tofacitinib dose of 5 or 10 mg twice daily without prior exposure to a different tofacitinib dose or adalimumab were included in the algorithm. Person-time with exposure to constant tofacitinib 5 or 10 mg twice daily would not sum to the overall exposure, because patients had exposure censored upon tofacitinib dose change.

Data collection, coding and adjudication

The safety database included patients receiving ≥1 tofacitinib dose.

Data were collected for all treatment-emergent AEs and serious AEs (SAEs) and coded using Medical Dictionary for Regulatory Activities (MedDRA) V.18.0. Details of comorbidities were requested at baseline.

An external, independent committee of infectious disease experts blindly adjudicated and classified all SIEs and all events of possible OIs occurring in the tofacitinib RA development programme. SIEs were defined as requiring hospitalisation or parenteral antimicrobial therapy, or otherwise meeting SAE criteria; patients with SIEs were discontinued from the study. HZ involving >2 adjacent dermatomes or with disseminated disease were considered OIs. TB screening was performed as previously described.20

Malignancy data were adjudicated as previously described.21

GI perforations were blindly adjudicated by two sponsor-independent reviewers (US board-certified practising gastroenterologists). SAEs in the clinical and safety databases that might reflect a GI perforation (see online supplementary appendix) were reviewed. Status was determined by two reviewers, with any differences resolved by a third reviewer. Clinical events reflecting an opening in the GI tract, including those associated with appendicitis and diverticulitis, were classified as confirmed GI perforations. Incidence rate (IR) of deaths occurring within 30 days after tofacitinib discontinuation was calculated.

Statistical analysis

Safety analyses were based on observed cases. Crude IRs per 100 patient-years were calculated for patients receiving tofacitinib and those in average or constant 5 or 10 mg twice daily groups. IRs for each AE were obtained by dividing the number of first-time occurrences of the AE over a time interval, by the total duration of study treatment exposure censored at time of first event, death or discontinuation from study time interval. An exact Poisson 95% CI adjusted for exposure time was calculated for IRs.

Standardised incidence ratios (SIRs) were calculated as the ratio of observed AEs to those in the US National Cancer Institute Surveillance and Epidemiology and End Results (SEER) database, 1992–201122; 95% CIs for SIRs were calculated following a Poisson distribution.

TB rates were stratified by geographical background rates using the WHO incidence categorisation of low, intermediate and high.23

To assess whether IRs increased over time, rates were examined within 6-month intervals of tofacitinib exposure. To investigate whether the hazard for developing a malignancy was constant over time, the probability distribution of time to first event was analysed using a Kaplan-Meier curve and cumulative hazard function. Patients were censored at their last day of tofacitinib exposure.

A Cox regression model evaluated risk factors for SIEs, HZ and OIs excluding TB; backward selection (stay fixed at 15%) was used to screen baseline factors: age, gender, region, body mass index (BMI), smoking history, RA disease duration, line of therapy, diabetes, chronic obstructive pulmonary disease (COPD), DAS-4(CRP), methotrexate use, glucocorticoid dose groups, absolute lymphocyte count (ALC), rheumatoid factor and Health Assessment Questionnaire-Disability Index (HAQ-DI) score to model the time to event from the first tofacitinib dose.


Patients and tofacitinib exposure

Six thousand one hundred and ninety-four patients received tofacitinib. Overall, 4794 (77.4%), 4032 (65.1%), 3351 (54.1%) and 2489 (40.2%) patients received tofacitinib for >12, 24, 36 and 48 months; overall median exposure was 3.38 patient-years. Demographics and disease characteristics were similar between groups (table 1).

Table 1

Patient demographics and baseline disease characteristics

AEs and SAEs

The most common AEs (all causality) were nasopharyngitis, upper respiratory tract infection and urinary tract infection (UTI); the most common System Organ Class of SAEs was infections and infestations.

IRs for AEs, discontinuations due to AEs, SAEs and deaths were similar for average and constant tofacitinib 5 and 10 mg twice daily (table 2). The most common causes of death were infections, cardiovascular events and malignancies.

Table 2

IRs (patients with events/100 patient-years; 95% CI) of AEs and SAEs (all-cause)

Serious infections

The most common types were pneumonia, HZ, UTI and cellulitis. IRs did not increase with longer treatment (figure 1A). IRs of SIEs and deaths due to infections by average and constant dose had overlapping 95% CIs (table 3).

Table 3

IRs of infections or malignancies, patients with events/100 patient-years (95% CI)

Figure 1

IRs for (A) SIE, (B) HZ (non-serious and serious) and (C) OI (excluding tuberculosis) over time for all tofacitinib doses. HZ, herpes zoster; IR, incidence rate; OI, opportunistic infection; SIE, serious infection event.

Based on Cox regression analysis, baseline glucocorticoid dose >0–<7.5 and ≥7.5 mg/day (selected based on clinical relevancy and sample size) were associated with increased risk of SIEs (HRs (95% CI) 1.6 (1.3 to 2.0) and 1.7 (1.3 to 2.2), respectively, vs no glucocorticoid use; p=0.0001) (figure 2A). Other significant baseline risk factors were higher age, presence of COPD, higher HAQ-DI score, higher BMI, prior confirmed post-baseline lymphopenia (<500 cells/µL), diabetes, female gender, line of therapy (3rd vs 2nd line), geographical region (Asia, Europe and Latin America, each vs US/Canada) and time-varying tofacitinib dose (referent to 5 mg twice daily) (all p<0.05) (figure 2A).

Figure 2

HRs of potential risk factors for events of serious infection (A), herpes zoster (B) and opportunistic infections excluding tuberculosis (C)—results from multivariable Cox regression models in the phases I–III and LTE studies *Medical history and/or complication of COPD. In Unit=x, ‘x’ is the change in the continuous variable corresponding to which the change in hazards is observed. Based on exposure period before lymphopenia <500 cells/µL versus exposure period after lymphopenia <500 cells/µL. COPD, chronic obstructive pulmonary disease; HAQ-DI, Health Assessment Questionnaire-Disability Index; LA, Latin America; LTE, long-term extension.

Six patients had lymphopenia <500 cells/µL, with a crude IR (95% CI) for SIEs of 8.3 (3.0 to 18.1), and 115 patients had lymphopenia ≥500–<1000 cells/µL, with a crude IR (95% CI) of 3.4 (2.8 to 4.1). Inclusion of confirmed ALC<500 cells/µL as a time-varying categorical covariate in a multivariable Cox regression model showed an increased risk of SIEs in the period following confirmed ALC <500 cells/µL (HR (95% CI) 2.5 (1.1 to 5.7)) versus the period before confirmed ALC <500 cells/µL. Evaluation of a threshold of ALC <1000 cells/µL (exposure period prior to lymphopenia <1000 cells/µL vs exposure period after lymphopenia <1000 cells/µL) showed a HR (95% CI) of 1.3 (1.0 to 1.6) (p=0.02), suggesting a trend towards increasing risk with lower lymphocyte counts. The 500 cells/µL threshold is recommended in the product label as the discontinuation criterion.

Herpes zoster

Overall, 703 patients developed HZ; IRs for the first occurrence of non-serious or serious HZ had overlapping 95% CIs for average and constant tofacitinib 5 and 10 mg twice daily (table 3). IR analysis by 6-month intervals did not reveal increasing IRs with longer exposure (figure 1B).

Most HZ cases (92%) involved one dermatome; IR (95% CI) of disseminated/multidermatomal HZ was 0.3 (0.2 to 0.4). Serious HZ was reported in 53 patients. HZ IRs (95% CI) were higher in Asia (5.9 (5.2 to 6.6)) than other regions (see online supplementary table S2).

Baseline glucocorticoid doses >0–<7.5 mg/day and ≥7.5 mg/day were associated with increased HZ (HR (95% CI) 1.5 (1.3 to 1.9) and 1.4 (1.1 to 1.8), respectively, vs no glucocorticoid use; p<0.0001) (figure 2B). Other significant risk factors were baseline age, geographical region, smoking history (ex-smoker and smoker, each vs never smoked) and time-varying tofacitinib dose (all p<0.05) (figure 2B).

Opportunistic infections

OIs excluding TB were reported in 61 patients (see online supplementary table S3), and OIs including TB in 97 patients; 95% CIs for IRs with average and constant tofacitinib 5 and 10 mg twice daily overlapped (table 3). IRs of OIs excluding TB did not increase with longer exposure (figure 1C). Thirty-one OI events were serious.

Baseline age, geographical region and time-varying tofacitinib dose were associated with increased OIs excluding TB (all p<0.05) (figure 2C).


Active TB was reported in 36 patients, four had latent TB at screening with a history of adequate treatment. TB IRs were similar for average and constant tofacitinib 5 and 10 mg twice daily (table 3). Pulmonary and non-pulmonary TB occurred in 17 and 19 patients, respectively. Most cases (28/36) occurred in geographical regions endemic for TB (see online supplementary table S4). At screening, 301 patients had latent TB in the phase I–III studies. Of these, 23 had untreated or inadequately treated latent TB, and were treated with isoniazid and permitted to enrol in the study after ≥1 month of treatment. None of these 301 patients developed active TB.


Malignancies (excluding non-melanoma skin cancer, NMSC) occurred in 173 patients and NMSC in 118 patients; analysis of IRs by dose revealed widely overlapping 95% CIs (table 3).

Geographical variation in the NMSC distribution was observed (see online supplementary table S5). Analyses of IRs by 6-month intervals did not reveal any trend (figure 3A, B). A constant hazard over time for developing a malignancy was seen before month 60 (see online supplementary figure S1). Estimation beyond month 60 was less precise due to small patient numbers and limited patient-years of exposure.

Figure 3

IRs for malignancies excluding NMSC (A) and (B) NMSC over time for all tofacitinib doses. IR, incidence rate; NMSC, non-melanoma skin cancer.

Age-adjusted and sex-adjusted SIR (95% CI) for all malignancies (excluding NMSC) versus SEER among tofacitinib-treated patients was 1.0 (0.8 to 1.1). SIRs (95% CI) for lymphoma, lung cancer and breast cancer were 2.6 (1.6 to 4.1), 1.4 (1.0 to 2.0) and 0.5 (0.3 to 0.7), respectively.

GI perforations

Twenty-two patients experienced GI perforations. IRs (95% CI) were 0.11 (0.07 to 0.17) overall (0.07 (0.02 to 0.17) and 0.14 (0.08 to 0.22) for average 5 and 10 mg twice daily; 0.00 (0.00 to 0.10) and 0.15 (0.07 to 0.27) for constant 5 and 10 mg twice daily). Perforations occurred in the large bowel, excluding anus and rectum (n=13), gastroduodenal area (n=3), small bowel (n=1), anus and rectum (n=2) and undetermined locations (n=3). All received concomitant therapy with non-steroidal anti-inflammatory drugs (NSAIDs) or corticosteroids. Ten patients received NSAIDs and corticosteroids; nine NSAIDs alone and three chronic corticosteroid therapy alone. Thirteen patients had a history of diverticulitis or diverticulosis and two additional patients had a history of gastric ulcers.


This analysis presents an integrated view of safety data across the tofacitinib RA development programme. Types and rates of AEs were similar to those observed in phase III trials, with no evidence of directional trends with longer-term tofacitinib exposure through 8.5 years.

SIEs are an identified risk with immunomodulatory medications in RA, including tofacitinib and biological DMARDs (bDMARDs). IRs were consistent with those reported in phase III tofacitinib trials.12–17 ,24 SIE IRs for bDMARDs in RA clinical trials range from 3.0 to 5.5 per 100 patient-years,25 and are similar to those reported with tumour necrosis factor inhibitors (TNFi) in RA registries (3.2–4.6 per 100 patient-years).26–28 IRs with tofacitinib were generally consistent with IRs with bDMARDs.25–28 Previous studies of registry data revealed a decrease in SIEs with TNFi over time, likely due to discontinuation in patients at increased risk of SIEs, and reduction in risk associated with improvement in function and decreased glucocorticoid use.29–31 In contrast, analyses of data from open-label LTE studies suggest that the SIE risk remains stable over time.32 ,33

Although an increase in SIEs was not detected here, the studies discontinued patients who developed SAEs, which may have depleted patients at risk of recurrent SAEs. Only time to first event, and not second and subsequent events were analysed.

Previous analyses of the tofacitinib RA development programme identified increased rates of HZ with tofacitinib versus placebo, with greater age and Asian locations identified as risk factors.34 Here, most HZ cases remained non-serious allowing study continuation; approximately 8% of patients with HZ experienced disseminated/multidermatomal HZ. HZ risk is elevated in patients with RA versus the general population;35 reports are conflicting regarding risk with bDMARDs.36 ,37 Rate comparisons between tofacitinib and bDMARDs are limited by lack of directly comparable data, but indirect comparisons indicate that IRs appear higher with tofacitinib versus other agents.37 However, rates of disseminated/multidermatomal HZ and serious HZ were low and comparable with rates of multidermatomal and ophthalmic HZ reported with bDMARDs or conventional DMARDs.37

Rates of fungal and viral OIs were similar to those previously reported,20 suggesting OIs are a risk with tofacitinib, although it is unclear whether risk differs from that associated with bDMARDs. Precise estimations of differences in OI risk among RA therapies are limited by heterogeneous definitions of OIs,38 geographical variability and lack of head-to-head studies with sufficient power to detect differences.

A higher incidence of TB has been observed in patients with RA versus the general population, and in those receiving TNFi.39 ,40 Here, the elevation in TB IR with tofacitinib was within the range for bDMARDs.33 ,41 TB rates reflected geographical background TB prevalence.

IR of malignancies (excluding NMSC) here was 0.89, similar to those observed in previous tofacitinib studies.18 ,21 Patients with RA are at higher risk of developing some malignancies than the general population.42 ,43 IRs and SIRs for malignancies (excluding NMSC) reported here are in the range reported with bDMARDs.33 ,44–46 Analysis of IR of malignancies by 6-month interval exposure revealed variability, whereas analysis of the probability distribution of time to first malignancy event revealed a constant hazard over time. Of note, no real dose difference was observed. Although SIRs for malignancies (excluding NMSC) by 6-month interval exposure were not evaluated, SIRs were stable over time in an analysis of malignancy data (to 10 April 2013) from 14 tofacitinib studies.21 Despite this, vigilance should be exercised when evaluating malignancy risk with long-term exposure.

Cardiovascular safety data for tofacitinib pooled from phase III and LTE studies (data cut-off: 10 April 2013) have been published.47 Similar findings were reported in this analysis of pooled data from phases I–III and LTE studies (data cut-off: 31 March 2015; see online supplementary appendix).

GI perforations are a known risk in patients with RA, especially in patients treated with NSAIDS or glucocorticoids.48 GI perforations with tocilizumab had an IR of 0.3 events/100 patient-years,49 and an observational study revealed an IR of 0.1 events/100 patient-years for TNFi.50 Another study showed higher GI perforation rates with bDMARDs and concomitant glucocorticoids (0.1 events/100 patient-years), versus bDMARDs without glucocorticoids (0.05 events/10 patient-years), indicating glucocorticoid use as a risk factor.51 GI perforations IRs here are within the range reported; most patients with GI perforations had underlying risk factors (eg, glucocorticoids and/or NSAIDs).

Comparison of IRs by dose is limited by several factors. One factor is the imbalance between the tofacitinib 5 and 10 mg twice daily doses in the LTE studies, due to the fact that patients were not randomised to treatment. Instead, patients from phase II and III studies were transitioned into the LTE on 5 and 10 mg twice daily, respectively, except for patients from China and Japan who initiated treatment with tofacitinib 5 mg twice daily per protocol. Therefore, differences in the chronology of LTE study initiation and patient numbers from phase II versus phase III led to a longer median duration of exposure with 5 mg twice daily, but higher overall patient-years' exposure with 10 mg twice daily. This, coupled with differences in geographical regions between trials, preclude definitive dose comparisons. This analysis is also limited by exclusion of patients upon development of a SAE and censoring at time of first event, meaning healthier patients remain at later time points. This limits our ability to evaluate potential changes in SAE rates over time with greater cumulative tofacitinib exposure. Furthermore, the average dosing approach used in the primary analysis did not consider the actual dose at the time of AE. As any method of dose categorisation in this population would have drawbacks, we used constant dosing in a sensitivity analysis to give a more complete picture. Comparisons with placebo were reported for the placebo-controlled phases II and III tofacitinib index studies;3–7 ,12–16 however, duration of treatment with placebo was short, and patient-years of exposure to placebo was limited, therefore we have not included placebo data in our analysis.

This report describes the tofacitinib safety profile across the RA clinical programme to 31 March 2015, with >6000 patients treated for ≤8.5 years. These data represent the most comprehensive view of long-term safety to date, and reveal a stable AE profile versus controlled studies and earlier analyses of LTE data. Ongoing comparative clinical studies and post-marketing surveillance will provide further information on the tofacitinib risk profile in the clinical trial and real-world settings.


Editorial support under the direction of the authors was provided by Kate Silverthorne on behalf of Complete Medical Communications and funded by Pfizer Inc. Pinaki Biswas, an employee of Pfizer Inc, provided statistical support for these analyses.


View Abstract


  • Handling editor Tore K Kvien

  • Contributors SBC made substantial contributions to the study conception and design and analysis and interpretation of the data; was involved in drafting the article and revising it critically for important intellectual content; and approved the final version to be published. YT, JRC, KK and RR made substantial contributions to the study conception and design, acquisition of data, and analysis and interpretation of the data; were involved in drafting the article and revising it critically for important intellectual content; and approved the final version to be published. XM made substantial contributions to the analysis and interpretation of the data; was involved in drafting the article and revising it critically for important intellectual content; and approved the final version to be published. EBL, CC-S, LW and JW made substantial contributions to acquisition of data and analysis and interpretation of the data; were involved in drafting the article and revising it critically for important intellectual content; and approved the final version to be published. PN made substantial contributions to acquisition of data; was involved in drafting the article and revising it critically for important intellectual content; and approved the final version to be published. KLW, KT, RDM and JG made substantial contributions to the analysis and interpretation of the data; were involved in drafting the article and revising it critically for important intellectual content; and approved the final version to be published.

  • Funding This study was sponsored by Pfizer Inc.

  • Competing interests SBC, KLW, CC-S and JW have served as consultants for, and have received speaker fees and honoraria from, Pfizer Inc. YT has served as a consultant for, and has received speaker fees and honoraria from, AbbVie, Asahi-kasei, Astellas Pharma, BMS, Chugai Pharma, Daiichi Sankyo, Eli Lilly, GlaxoSmithKline, Janssen, Mitsubishi-Tanabe, Pfizer Inc, Sanofi, Takeda, Teijin and YL Biologics. XM has served as a consultant for, and has received speaker fees and honoraria from, BMS, GlaxoSmithKline, Pfizer Inc and UCB. EBL has served as a consultant for Pfizer Inc. KT, RDM, JG, KK and LW are employees and shareholders of Pfizer Inc. RR was an employee of Pfizer Inc at the time these analyses were conducted and holds stock/stock options in Pfizer Inc.

  • Patient consent Obtained.

  • Ethics approval Multiple Ethics Committees/Institutional Review Boards approved the studies. Additional details available upon request.

  • 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.