Article Text

Download PDFPDF

End-tidal and arterial carbon dioxide gradient in serious traumatic brain injury after prehospital emergency anaesthesia: a retrospective observational study
  1. James Price1,2,
  2. Daniel D Sandbach1,
  3. Ari Ercole1,3,
  4. Alastair Wilson1,4,
  5. Ed Benjamin Graham Barnard1,2,5
  1. 1 Department of Research, Audit, Innovation, & Development (RAID), East Anglian Air Ambulance, Norwich, UK
  2. 2 Emergency Department, Addenbrooke's Hospital, Cambridge, UK
  3. 3 University of Cambridge Division of Anaesthesia, Addenbrooke's Hospital, Cambridge, UK
  4. 4 Emergency Department (Retired), Royal London Hospital, London, UK
  5. 5 Academic Department of Military Emergency Medicine, Royal Centre for Defence Medicine (Research & Academia), Birmingham, UK
  1. Correspondence to James Price, Department of Research, Audit, Innovation, & Development (RAID), East Anglian Air Ambulance, Norwich, UK; james.price{at}eaaa.org.uk

Abstract

Objectives In the UK, 20% of patients with severe traumatic brain injury (TBI) receive prehospital emergency anaesthesia (PHEA). Current guidance recommends an end-tidal carbon dioxide (ETCO2) of 4.0–4.5 kPa (30.0–33.8 mm Hg) to achieve a low-normal arterial partial pressure of CO2 (PaCO2), and reduce secondary brain injury. This recommendation assumes a 0.5 kPa (3.8 mm Hg) ETCO2–PaCO2 gradient. However, the gradient in the acute phase of TBI is unknown. The primary aim was to report the ETCO2–PaCO2 gradient of TBI patients at hospital arrival.

Methods A retrospective cohort study of adult patients with serious TBI, who received a PHEA by a prehospital critical care team in the East of England between 1 April 2015 and 31 December 2017. Linear regression was performed to test for correlation and reported as R-squared (R2). A Bland-Altman plot was used to test for paired ETCO2 and PaCO2 agreement and reported with 95% CI. ETCO2–PaCO2 gradient data were compared with a two-tailed, unpaired, t-test.

Results 107 patients were eligible for inclusion. Sixty-seven patients did not receive a PaCO2 sample within 30 min of hospital arrival and were therefore excluded. Forty patients had complete data and were included in the final analysis; per protocol. The mean ETCO2–PaCO2 gradient was 1.7 (±1.0) kPa (12.8 mm Hg), with moderate correlation (R2=0.23, p=0.002). The Bland-Altman bias was 1.7 (95% CI 1.4 to 2.0) kPa with upper and lower limits of agreement of 3.6 (95% CI 3.0 to 4.1) kPa and −0.2 (95% CI −0.8 to 0.3) kPa, respectively. There was no evidence of a larger gradient in more severe TBI (p=0.29). There was no significant gradient correlation in patients with a coexisting serious thoracic injury (R2=0.13, p=0.10), and this cohort had a larger ETCO2–PaCO2 gradient, 2.0 (±1.1) kPa (15.1 mm Hg), p=0.01. Patients who underwent prehospital arterial blood sampling had an arrival PaCO2 of 4.7 (±0.2) kPa (35.1 mm Hg).

Conclusion There is only moderate correlation of ETCO2 and PaCO2 at hospital arrival in patients with serious TBI. The mean ETCO2–PaCO2 gradient was 1.7 (±1.0) kPa (12.8 mm Hg). Lower ETCO2 targets than previously recommended may be safe and appropriate, and there may be a role for prehospital PaCO2 measurement.

  • prehospital care
  • anaesthesia
  • trauma
  • head
http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

View Full Text

Statistics from Altmetric.com

Key messages

What is already known on this subject

  • The optimal management of ventilation in patients with serious traumatic brain injury (TBI) reduces secondary brain insult.

  • Current guidance recommends the use of end-tidal carbon dioxide (ETCO2) as a surrogate for arterial carbon dioxide (PaCO2) to guide ventilation strategy in the prehospital environment.

  • The effect of the ETCO2–PaCO2 gradient on resultant PaCO2 early in this disease process is unknown.

What this study adds

  • This retrospective cohort study of 40 patients receiving prehospital emergency anaesthesia found only moderate correlation of ETCO2 and PaCO2 in patients with serious TBI, thus the use of ETCO2 to guide ventilation is a blunt tool in this cohort.

  • The ETCO2–PaCO2 gradient is larger than previously reported and is increased by the presence of coexisting thoracic injuries. Prehospital PaCO2 measurement resulted in more accurate titration of ventilatory strategy.

Background

Traumatic brain injury (TBI) is the leading cause of death and disability following trauma, with 69 million estimated new cases each year worldwide.1 Optimising cerebral blood flow (CBF) is the mainstay of treatment to prevent secondary brain injury and reduce mortality.2 Early management of raised intracranial pressure (ICP) includes controlled ventilation via prehospital emergency anaesthesia (PHEA).3 A growing number of emergency medical services are providing this.4 At present, it is estimated that one in five patients with severe TBI undergo PHEA.5

The use of end-tidal capnography to confirm endotracheal tube position is widely accepted.6 Current guidance encourages the use of end-tidal carbon dioxide (ETCO2) as a surrogate for the arterial partial pressure of carbon dioxide (PaCO2) to guide ventilation.7 8 Under normal physiological conditions, alveolar dead space accounts for a 0.5 kPa (3.8 mm Hg) ETCO2–PaCO2 gradient in ventilated patients.9 In-hospital data demonstrate good correlation in patients with TBI.10–12 The Association of Anaesthetists recommend an ETCO2 target of 4.0–4.5 kPa (30.0–33.8 mm Hg) to achieve normocapnia.7

The mechanisms of autoregulation of CBF differ with changes in ICP.13 Respiratory dysfunction and loss of autoregulation is common in patients with TBI,14 and a 1% change in PaCO2 can lead to changes of up to 4% in CBF.15 Prehospital data evaluating the use of ETCO2 are limited, but are not consistent with in-hospital findings; leading to some criticism of the use of ETCO2 to guide ventilation.16–18 In most cases of TBI, an increase in ICP evolves over hours, and therefore moderate hypercapnia prehospital would be expected to increase CBF. However, the effect of the ETCO2–PaCO2 gradient on resultant PaCO2 early in this disease process is unknown. The primary objective of this study was to report the ETCO2–PaCO2 gradient at hospital arrival in a cohort of patients with serious TBI who underwent PHEA.

Hypothesis

Patients with serious TBI who undergo PHEA have an ETCO2–PaCO2 gradient greater than 0.5 kPa (3.8 mm Hg).

Methods

In this retrospective observational study, a convenience sample of patients who underwent PHEA by a single UK prehospital team (East Anglian Air Ambulance, EAAA) between 1 April 2015 and 31 December 2017 was extracted from the EAAA electronic medical record.

Prehospital critical care team

EAAA is a medical charity that provides prehospital critical care to the statutory ambulance service in the East of England (East of England Ambulance Service NHS Trust). EAAA operates from two bases (Cambridge and Norwich), dispatching a physician-paramedic prehospital critical care team in either an H145 helicopter or rapid response vehicle, depending on patient location, weather constraints and time of day (H145 available 07:00–23:59 in Cambridge, and 07:00–19:00 in Norwich). During the study period, the Cambridge EAAA base was operational from 07:00 to 01:30 daily, and the Norwich base was operational 24 hours per day 7 days per week with a paramedic-only service (without PHEA capability) from 19:00 through 07:00.

Inclusion criteria

In order to report outcomes, only patients who were primarily transported by EAAA to the regional neurosciences (and major trauma) centre (Cambridge University Hospitals NHS Foundation Trust, CUH) were included. This allowed cross-reference of injury patterns and outcomes collated by the CUH Trauma Office, and also for retrieval of PaCO2 data from the CUH electronic medical record. Patients were included if they were attended by EAAA, were ≥18 years old, underwent PHEA, were transported to CUH from scene, had a serious (or more severe) TBI and had complete data. Consistent with recent large data methodology, patients without a PaCO2 measurement within 30 min of hospital arrival were excluded.19

Definitions

PHEA was defined as drug-assisted endotracheal intubation in the prehospital setting. Serious TBI was defined as a retrospectively applied Abbreviated Injury Scale (AIS) score for ‘head’≥3, and serious thoracic injury was defined as an AIS ≥3 for ‘thorax’.

Data collection

The side-stream ETCO2 values and BP (reported as mean arterial pressure, MAP) at hospital handover were obtained from the EAAA electronic medical record that includes a time-stamped download from the prehospital monitor (ZOLL X Series Monitor/Defibrillator, ZOLL Medical Corporation of Asahi Kasei Corp., Tokyo). The in-hospital PaCO2 values were obtained from the CUH electronic medical record that includes a time-stamped download of data from an ABG analyzer (COBAS B 221 Blood Gas Analyzer, Roche Diagnostics, Indianapolis, IN, USA). ETCO2 and PaCO2 data were recorded in kilopascal (kPa) units; a conversion of (kPa*7.50062=mm Hg) has been used to present units of millimetres of mercury alongside kPa.

Demographic, mechanism of injury, injury severity (AIS and Injury Severity Score (ISS)), 30 day mortality and functional outcome (Glasgow Outcome Scale (GOS) score) data were obtained from the CUH Trauma Office records.

Primary outcome

The primary outcome was to report the ETCO2–PaCO2 gradient at hospital arrival.

Secondary outcomes

The secondary outcomes were to report the relationship between the severity of TBI (serious (AIS=3) and severe (AIS=4) versus critical (AIS=5)) and ETCO2–PaCO2 gradient at hospital arrival; to report the effect of a coexisting serious (AIS ≥3) thoracic injury on the ETCO2–PaCO2 gradient at hospital arrival; and to compare the PaCO2 at hospital arrival in patients that received prehospital arterial blood sampling.

Statistical analysis

Basic demographic, mechanism of injury and injury data have been reported as number (percentage) and mean (±SD) or median (IQR) as appropriate. ETCO2 and PaCO2 data have been reported as percentage (95% CI) and mean (±SD). Comparisons of unpaired, normally distributed, continuous variables were undertaken with a two-tailed unpaired t-test (with Welch’s correction if samples had unequal deviations).

Fisher’s exact test has been used to compare proportions. Linear regressions have been performed to test for correlation and are reported as R-squared (R2) with gradient of the slope (m). A Bland-Altman plot has been used to test for agreement between paired ETCO2 and PaCO2 data, and has been reported as bias (95% CI) with upper and lower limits of agreement.

Statistical analyses were performed using the R statistical programming language (R Core Team (2018); R: A language and environment for statistical computing (R Foundation for Statistical Computing, Vienna, Austria); significance was predefined at <0.05 and no corrections were made for multiple comparisons.

Patient and public involvement

Patients or the public were not involved in the design, conduct, reporting or dissemination plans of our research.

Results

Demographics and injury

107 patients were eligible for inclusion. Sixty-seven patients did not receive a PaCO2 sample within 30 min of hospital arrival and were therefore excluded. Forty patients had complete data and were included in the final analysis; per protocol. The median age was 45 (23–63) years, 24 (60.0%) were male. The most prevalent mechanism of injury was road traffic collision, n=22 (55.0%). The median ISS was 31 (26–38), and 97.5% of the cohort had an ISS ≥15. The mean hospital arrival MAP was 93.7 (±24.4) mm Hg. Overall, 30 day mortality was 30.0% (95% CI 16.6 to 47.9), and all survivors had a GOS ≥3 (severe disability or better), table 1.

Table 1

Study population demographics

ETCO2–PaCO2

Overall, the mean ETCO2 and PaCO2 were 4.1 (±0.7) kPa (30.7 mm Hg) and 5.8 (±1.1) kPa (43.3 mm Hg), respectively. In 38/40 (95.0%, 95% CI 83.5 to 98.6) patients, the PaCO2 was more than 0.5 kPa (3.8 mm Hg) higher than the ETCO2. No patients were hypocapnic (PaCO2 lower than 4.0 kPa (30.0 mm Hg)) and there was moderate correlation between ETCO2 and PaCO2 (R2=0.23, m=0.72, p=0.002), figure 1.

Figure 1

Scatter plot of the relationship between ETCO2 and PaCO2 at hospital arrival. 1.0kPa=7.50062 mm Hg. The ETCO2–PaCO2 gradient was 1.7 (±1.0) kPa (12.8 mm Hg). The mean is plotted as a single line and SD as dashed lines. ETCO2, end-tidal carbon dioxide; PaCO2, arterial partial pressure of carbon dioxide.

Primary outcome

The mean ETCO2–PaCO2 gradient was 1.7 (±1.0) kPa (12.8 mm Hg), figure 1. The Bland-Altman analysis looking at the difference between PaCO2 and ETCO2 showed the bias to be 1.7 (95% CI 1.4 to 2.0) kPa with upper and lower limits of agreement of 3.6 (95% CI 3.0 to 4.1) kPa and −0.2 (95% CI −0.8 to 0.3) kPa, respectively, figure 2.

Figure 2

Bland-Altman plot of the agreement between ETCO2 and PaCO2 at hospital arrival. 1.0kPa=7.50062 mm Hg. The bias is 1.7 (95% CI 1.4 to 2.0) kPa with upper and lower limits of agreement of 3.6 (95% CI 3.0 to 4.1) kPa and −0.2 (95% CI −0.8 to 0.3) kPa, respectively. The bias is plotted as a single line, and the upper and lower limits of agreement as dashed lines. ETCO2, end-tidal carbon dioxide; PaCO2, arterial partial pressure of carbon dioxide.

Secondary outcomes

Severity of TBI

Patients with a critical TBI (AIS of 5), n=24 (60.0%) had a comparable mean ETCO2–PaCO2 gradient compared with those with a serious or severe (AIS=3 or 4) injury—1.8 (±1.1) kPa (13.5 mm Hg) and 1.5 (±0.7) kPa (11.2 mm Hg), respectively, p=0.29. Both of these groups (AIS=5 and AIS=3 or 4) demonstrated a moderate correlation (R2=0.22, m=0.64, p=0.02 and R2=0.31, m=1.31, p=0.03, respectively).

Thoracic injury

22/40 (55.0%) patients had a serious thoracic injury. The mean ETCO2–PaCO2 gradient in this subgroup was 2.0 (±1.1) kPa (15.1 mm Hg), significantly higher than the 18/40 patients without a serious thoracic injury—1.3 (±0.5) kPa (9.5 mm Hg), p=0.01. There was no significant correlation in the thoracic injury subgroup (R2=0.13, m=0.54, p=0.10), but there was a significant correlation in the group without serious thoracic injury (R2=0.51, m=0.78, p<0.001). There was no significant difference in mean MAP at hospital arrival between patients with and without a serious thoracic injury—91.9 (±25.3) mm Hg and 95.9 (±23.7) mm Hg, respectively, p=0.60.

Prehospital PaCO2 measurement

Seven (17.5%) patients underwent prehospital measurement of PaCO2, using the iSTAT (Abbott Laboratories, Illinois, USA) point-of-care ABG (POC-ABG) analyzer. The POC-ABG group had a significantly lower, and more appropriate hospital arrival PaCO2, compared with the no-ABG group—4.7 (±0.2) kPa (35.1 mm Hg) versus 6.0 (±1.1) kPa (45.0 mm Hg), respectively, p=0.7×10−7.

Discussion

This study has demonstrated that patients with an AIS ≥3 TBI have a larger ETCO2–PaCO2 gradient at hospital arrival than previously reported. The severity of TBI was not significantly associated with the ETCO2–PaCO2 gradient. However, the presence of a coexisting serious thoracic injury was associated with a significantly larger ETCO2–PaCO2 gradient, which is manifest as an apparent bias in the Bland-Altman plot of figure 2. In the serious thoracic injury cohort, there was no evidence of correlation between ETCO2 and PaCO2. All patients who underwent prehospital PaCO2 measurement had a low-normal PaCO2 at hospital arrival.

PaCO2 is a major determinant of CBF through its effects on cerebral vascular tone. Ventilatory control is crucial to the management of TBI.20 Even modest hypercapnia may result in substantial increases in ICP when intracranial compliance is poor.21 Conversely, even modest hyperventilation has been shown to lead to dangerous cerebral ischaemia,22 and this may be of particular importance in the first hours after TBI when hypoperfusion is a dominant pathology.23

Current guidance recommends an ETCO2 of 4.0–4.5 kPa (30.0–33.8 mm Hg) as a surrogate for a low-normal PaCO2.7 This guidance relies on the assumption that the ETCO2–PaCO2 gradient is approximately 0.5 kPa (3.8 mm Hg), and is predominantly extrapolated from healthy individuals. In this study, the mean hospital arrival ETCO2 was 4.1 (±0.7) kPa (30.7 mm Hg), which suggests that the prehospital providers adhered closely to the extant guidelines. Despite this, the mean PaCO2 was 5.8 (±1.1) kPa (43.3 mm Hg)—far in excess of the target of 4.5–5.0 kPa,24 owing to a mean ETCO2–PaCO2 gradient of 1.7 (±1.0) kPa (12.8 mm Hg). The result of this is that when relying on ETCO2 as a surrogate, providers may not achieve an optimal prehospital PaCO2.

The arterial-alveolar CO2 gradient is determined by the ratio of physiological dead space and tidal volume. Since 2000 and the publication of the ARDSNet study, ventilatory practice has shifted towards the use of lower tidal volumes.25 Therefore, in contemporary practice the ETCO2–PaCO2 gradient would be expected to be larger because the dead space is more appreciable, and it may be that the target ETCO2 needs to be reconsidered. It is tempting to advocate a lower prehospital ETCO2—perhaps 2.8–3.3 kPa (21.0–24.8 mm Hg) in patients suspected of having an AIS ≥3 TBI (assuming a gradient of 1.7 kPa (12.8 mm Hg)), particularly in the setting of coexisting thoracic injury. However, this strategy may risk hypocapnia (PaCO2 <4.0 kPa (30.0 mm Hg)), which in the setting of a very high ICP may be of benefit via hypocapnic arterial vasoconstriction,21 but in the more likely prehospital clinical scenario (<24 hours after TBI) of normal ICP, may lead to severe cerebral ischaemia and worse outcomes.26

There was only moderate correlation between ETCO2 and PaCO2; statistically, only 23% of the variance (R2, the coefficient of determination) observed in ETCO2 can be explained by the PaCO2. There was no ETCO2–PaCO2 gradient correlation in the subgroup with an AIS ≥3 thoracic injury. We presume that this is owing to the effect of heterogeneous physiological dead space (caused by the thoracic injury) on the arterial-alveolar CO2 gradient. Therefore, in those without an AIS ≥3 thoracic injury the ETCO2–PaCO2 gradient correlation was better; statistically 51% of the variance observed in ETCO2 can be explained by PaCO2. In order to ensure that the increased gradient observed in the AIS ≥3 thoracic injury group was not simply due to a lower systemic perfusion compared with the group without serious thoracic injury, MAP at hospital arrival was reported between the two groups. Although there is a known relationship between tissue perfusion and ETCO2,11 there was no difference between the MAP (used as a surrogate for perfusion) at hospital arrival between those with and without an AIS ≥3 thoracic injury. We think that this strengthens the theory that thoracic injury increases the physiological dead space. It is unclear from these data what variable(s) make up the remaining 49% of variance observed. We acknowledge the transfer from a prehospital to an in-hospital ventilator may be one factor in this. However, EAAA and CUH use identical ventilators (Dräger Oxylog 3000, Drägerwerk AG &Co., Lübeck, Germany), and it is standard practice to commence in-hospital ventilation using the prehospital settings.

It is not necessarily surprising that in this small sample with significant variance we were not able to demonstrate a significant difference between the severity of TBI (as measured by AIS) and the ETCO2–PaCO2 gradient, and it is possible that this represents a type-2 error. Previous work has demonstrated an increased gradient in more severe injury (higher ISS), but has not examined head or thoracic injury specifically,16 and has shown an increased mortality in ‘abnormal’ hospital-arrival ETCO2.27 In the subgroup of patients with a serious thoracic injury, the ETCO2–PaCO2 gradient was significantly higher than those without a serious thoracic injury, further compounding the inaccuracy in estimating PaCO2 from ETCO2 in the severely injured trauma patient. The patients in our study were all primarily transported to the regional neurosciences (and major trauma) centre by a prehospital critical care team. Therefore, it is possible that we are missing data from both ends of the severity spectrum: patients with a lesser injury and those with a more immediate requirement for in-hospital resuscitation may have been initially transported to a local trauma unit hospital. Our limited analysis of the effect of severity of TBI on gradient demonstrated a non-significant trend of increasing gradient with increasing TBI severity. However, the pragmatic clinical benefits of knowing if severity of TBI affects the ETCO2–PaCO2 gradient is minimal—without good evidence of strong correlation between ETCO2 and PaCO2, the provider cannot rely on ETCO2 as a surrogate.

Inadvertent hypocapnia is an important factor in avoidable neuronal injury following TBI via a reduction in CBF and an increase in cerebral oxygen consumption.22 While we have concentrated on hypercapnia in this study, it is encouraging to see that even in the presence of significant heterogeneity in ETCO2–PaCO2 gradient in this population, no patients were hypocapnic on arrival to hospital using the extant guidelines.

Even from this small sample of patients it is evident that using the ETCO2 as a surrogate for PaCO2 following TBI is a blunt tool. There is a larger than previously reported mean ETCO2–PaCO2 gradient and only moderate correlation. The practice of low tidal volume ventilation may necessitate a reconsideration of the expected ETCO2–PaCO2 gradient, but the lack of correlation in those with a concomitant thoracic injury means that significant inaccuracy in using ETCO2 as a surrogate for PaCO2 would likely continue. Numerous other surrogates have been used for PaCO2, including transcutaneous CO2 monitoring and capillary blood gas analysis.28 29 However, at present, an ABG sample is the only reliable way to obtain an accurate PaCO2 with which to guide ventilation in TBI. The advancement in technology and production of smaller, portable analyzers allow providers to accurately determine PaCO2 prehospital and in other resource-limited settings.30 31 A small number of patients in our study underwent prehospital measurement of PaCO2 using a POC analyzer. There was evidence during chart review of these seven patients that the providers made alterations to the ventilation strategy after PaCO2 analysis; they all demonstrated a favourable hospital arrival PaCO2.

This study used data from a single prehospital critical care service and a single regional neurosciences centre, and as such the results may not be widely applicable. In order to increase the number of patients in this study, we included patients who had an ABG up to 30 min after hospital arrival. It is therefore possible that a proportion of the gradient variance observed was due to changes in ventilation strategy during this period, but standard clinical practice means this should be clinically negligible. The availability of these data very early in this disease process is extremely limited, resulting in a small sample size. It is possible that non-significant findings are due to a lack of statistical power.

Conclusion

There is only moderate correlation of ETCO2 and PaCO2 at hospital arrival in patients with serious TBI. The mean ETCO2–PaCO2 gradient was 1.7 (±1.0) kPa (12.8 mm Hg)—greater than previously reported. Lower ETCO2 targets than previously recommended may be safe and appropriate, particularly in the presence of thoracic injury. There may be a role for prehospital PaCO2 measurement.

Acknowledgments

We acknowledge the assistance of Assiah Mahmood and Jacques Bowman of the CUH Trauma Office in compiling the original data.

References

View Abstract

Footnotes

  • Handling editor Ellen J Weber

  • Twitter @edbarn

  • Contributors The study was conceived by JP, DDS and EBGB. The study permissions were obtained by EBGB and AW. Data acquisition was undertaken by JP and DDS. DDS, AE and EBGB interpreted the data. The manuscript was prepared by JP and EBGB. Critical revisions were done by AE, AW and EBGB. All authors reviewed and approved the final draft.

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

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting or dissemination plans of this research.

  • Patient consent for publication Not required.

  • Ethics approval This service evaluation was registered with both EAAA and CUH. Local agreement for the use of anonymised data from the EAAA electronic medical record was granted through extant data use protocols. Anonymised, linked data were obtained from the CUH Trauma Office, and PaCO2 values obtained from the CUH electronic medical record. Ethical review was undertaken by the Cambridge University Hospitals NHS Foundation Trust Safety and Quality Support Department (reference: PRN7866).

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

  • Data availability statement Data are available upon reasonable request.

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.