Introduction
The final stage in palliation of patients with functional single ventricle CHD is total cavopulmonary anastomosis (Fontan/Kreutzer operation). This palliative procedure creates a “bottleneck” in the pulmonary circulation, and overall cardiac output is limited by the amount of pulmonary blood flow and resistance in the circuit. Reference Gewillig and Brown1 The mean pulmonary artery pressures or Fontan pressures is an indirect measure of central venous pressure required to maintain this circulation. Reference Gewillig and Brown1,Reference Venna, Deshpande, Downing, John and dUdekem2 Haemodynamic assessment of Fontan pressures during cardiac catheterisation reflects the pressures in the pulmonary circulation during the resting state. This data, although useful, does not allow assessment of the Fontan pressures in response to a variety of physiological stressors, such as exercise, illness, or change in altitude. Resting Fontan pressures may not predict Fontan failure or Fontan-related complications, necessitating more dynamic Fontan pressure easurements. Reference Venna, Deshpande, Downing, John and dUdekem2,Reference Bulut, Bailey and Bhat3.
Therefore, implantable haemodynamic monitors (IHM, CardioMEMs HF system) present a unique and attractive opportunity for remote monitoring of Fontan circulation. Starting in September 2021, our Fontan team expanded the use of this device in paediatric Fontan patients with failing Fontan physiology. We have previously demonstrated the safety of this device in paediatric Fontan patients as young as 9 years of age; Reference Bhat, Graziano, Garn and Franklin5 while other researchers have reported utility of implantable haemodynamic monitor in adult Fontan patients. Reference Bradley, Jassal, Moore-Clingenpeel, Abraham, Berman and Daniels6–Reference Salavitabar, Bradley and Chisolm8 We have also shown the potential role of implantable haemodynamic monitor for assessing dynamic changes in Fontan pressures during exercise in paediatric patients with Fontan physiology. Reference Sabati, Chesney, Montchal and Bhat9 Besides exercise, exposure to higher altitude may also affect Fontan pressure. However, the true extent of this effect on Fontan circulation is uncertain. Reference Zhou, Malhotra and Yu10–Reference Malhotra, Ivy and Mitchell16 Some researchers have attempted direct measurement of Fontan pressure using invasive central venous pressure monitoring to assess the effect of altitude. Reference Müller, Härtel and Schmitz17
Implantable haemodynamic monitors (IHM, CardioMEMS HF system) have been used successfully for management of heart failure in adults. Reference Zile, Mehra and Ducharme4 The implantable haemodynamic monitor device is originally designed for the pulsatile two-ventricle circulation, and the readings are adapted for non-pulsatile Fontan circulation, which represents an off-label use of this product. CardioMEMS HF is a wireless, battery-free, pulmonary artery pressures monitoring system (Abbott) that is concurrently powered and interrogated via an external antenna. When the sensor experiences the pressure from the arterial system, it responds by creating deflections of the pressure-sensitive surface, thereby causing a shift in the resonant frequency. Reference Abraham, Adamson and Bourge22 This change in frequency is coupled electromagnetically by an external antenna (external measuring unit), which is held against the patient’s body or embedded in a pillow. Reference Abraham, Adamson and Bourge22 The implantable haemodynamic monitor sensors are intended to operate in the radiofrequency range 30–37.5 MHz as a result of internal sensor characteristics and current atmospheric conditions pre-implant, and those operating above 37.5 MHz have an increased rate of inaccuracy complaints and readings identified as suspect, and that altitude can alter this.
There are a few isolated reports of significant change in the implantable haemodynamic monitor readings when adult heart failure patients travelled to higher altitudes. Reference Letourneau, Brancheau, Estes and Zughaib18,Reference Hamadeh, Milligan and Minniefield19 The CardioMEMS user manual (2022) mentioned potential inaccuracies in the device when readings are taken over ∼2,000 feet (610 m) above sea level. Reference Boltz20 A formal sensor-device correction notification was released in February 2023, advising instructions for use of an altitude-correction factor of + 2 mmHg for every 1000 ft (305m) increase in altitude from the level at initial calibration. 21 There is limited experience with use of implantable haemodynamic monitor in the Fontan circulation, with additionally limitations and challenges in evaluating the performance of the sensor when used in young Fontan patients, living in remote, rural, and high-altitude areas with limited access to care.
Our tertiary referral centre has a unique population covering a large part of rural, higher-altitude Arizona, with limited access to care. This has allowed our team a unique experience of using the implantable haemodynamic monitor in patients from diverse geographical locations within the state. Applying this device in this setting has presented unique challenges, and we describe our centre’s experience with the performance of the device and potential solutions for improving device accuracy in this group.
Methods
This was a retrospective chart review study approved by our institutional review board. All patients with Fontan circulation who underwent implantable haemodynamic monitor placement were included. Implantable haemodynamic monitor readings were recorded at the implantation centre, prior to discharge (labelled, “predischarge reading”). Parents/caregivers were instructed to perform home device readings within 24 hours, and an average of 3 readings over the next one week were recorded (labelled “home reading”). Demographic characteristics, including the home altitude, distance from referral centre, age at implant, gender, underlying cardiac diagnosis, age at Fontan completion, type of Fontan, medication, and haemodynamic data, at the time of cardiac catheterisation, including presence of any additional metallic hardware, were collected, along with any significant Fontan-related complications. Sensor readings were analysed for occurrence of measurement discrepancy, which was defined as non-physiological, negative, or significantly lower reading than baseline (reading at the time of implantation). Continuous variables were represented in mean and standard error of mean. Paired t-test was applied to compare Fontan pressures assessed on the same patient using the implantable haemodynamic monitor and cardiac catheterisation. Categorical variables were compared using Fischer’s exact test and continuous variables were compared using Student t-test. A p value of less than 0.05 was considered significant.
Technical considerations
Implantable haemodynamic monitor was implanted in a distal branch of the left pulmonary artery in all patients under fluoroscopic guidance during haemodynamic cardiac catheterisation under general anaesthesia. Reference Bhat, Graziano, Garn and Franklin5 The reason for selecting an inferior and lateral branch of left pulmonary artery for implantation is based on current recommendations for implantable haemodynamic monitor. Reference Bärtsch and Gibbs23 As the left pulmonary artery courses posteriorly after it takes off, this branch is preferred as the sensor can be interrogated from the back of the patient more effectively. The left pulmonary artery was accessed inferiorly via the Fontan conduit or superiorly via the superior cavopulmonary connection. Venous access was obtained typically with a 7 Fr balloon wedge catheter, over 11 mm sheath, and the left pulmonary artery was engaged using a 0.035˝ angled Glide wire. We determined the target implant site in a branch of the left pulmonary artery, measuring at least 7 mm and up to 15 mm in diameter, and placed the Glidewire tip approximately 5 cm beyond the target implant site (Figure 1). A pulmonary angiogram was obtained in the antero-posterior and lateral planes. We flushed the port of the sensor delivery catheter with heparinised saline and allowed 10–15 s of agitation with saline to activate the hydrophilic coating on the delivery catheter. We used fluoroscopic guidance to advance and release the sensor at the target location. Two patients had a stent in the left pulmonary artery, and one patient had a stent in their innominate vein at the time of implant. We did not experience any technical difficulty placing the device in patients as young as 9 years of age, and even in a few patients who had a pre-existing stent in their proximal left pulmonary artery.
(a) Technical considerations while implanting IHM in a Fontan patient. The left panel shows the target implant site in the branch left pulmonary artery. The right panel shows the fluoroscopic view of the implanted IHM in a Fontan patient with a pre-existing stent in the left pulmonary artery. ( b ) shows examples of sensor readings from Fontan patients.
Device calibration and altitude correction
During cardiac catheterisation, we inserted the balloon wedge catheter to measure the mean pulmonary artery pressure, and this waveform was applied to calibrate the device sensor. The Implantable haemodynamic monitor unit consists of two different interrogating units with external antennae; the hospital electronic unit shaped like a wand; and a home unit shaped like a pillow. For calibration, we placed the hospital electronics unit wand behind the patient and assessed signal strength. We used an average of 3 to 5 mean pulmonary artery pressures consecutive readings to calibrate the sensor. After the procedure, as the patients recovered, another reading was obtained using the hospital unit (wand) prior to discharge (pre-discharge reading). Patients were educated and advised to use the home unit (pillow) to interrogate the device within 24 hours of returning home and several times during the first week (home reading). We confirmed the readings with the family during a 1-week telehealth follow-up. Home readings were obtained by the caregiver/patient with the patient lying supine. Clinic readings were obtained using the hospital electronics unit (wand) with the patient lying down to prevent any discrepancy due to postural change.
The sensor was calibrated with the mean pulmonary artery pressure obtained in our cardiac catheterisation laboratory (located ∼1000 ft above sea level). Our second patient (patient 2) had device implantation in January 2022 and lived at higher altitude (∼7000 feet above sea level), and we noted a sharp increase in the device recordings when she returned home and drop in the device readings while travelling to lower altitude. We contacted the device team and followed their recommendation of adjusting the device readings per manufacturer recommended altitude correction factor (CardioMEMS user manual, 2022). 21 In patients who lived at an altitude above 2000 feet above sea level, we report the actual and adjusted device readings for change in altitude (Table 2), for example, if the patient lives at 4000 feet, and their home device reading was 20mmHg; device recording was adjusted for altitude change (change in altitude 4000 ft – 1000 ft = 3000 ft; hence, expected error of 2 mm/1000 ft = 2 × 3 = 6 mmHg, adjusted home reading: 20 mmHg – 6mmHg = 14 mmHg). First week home readings were an average of several readings that patients obtained during the first week after returning home. Patients were advised to alert the Fontan team for any change in their residence. During follow-up, the same patient (patient 2) relocated from higher altitude to lower altitude, and subsequent data was reported without the correction factor.
Anticoagulation strategy
Most patients were on aspirin prophylactic dose (81 mg daily) prior to implant, and the medication was held 5 days prior to the implant date. Patients were heparinised during the implantation. Post procedure, patients were placed on oral Direct oral anticoagulant (DOAC) agent (Rivaroxaban), and aspirin was continued for 3 months with some exceptions. Patient 6 complained of easy bruising after the procedure, hence, aspirin was discontinued, and she was maintained on DOAC only. Three patients had severe, recurrent protein-losing enteropathy and were on DOACs prior to procedure (patients 9,11, and 16). These patients were maintained on DOAC after the procedure, for long-term thromboprophylaxis. During the follow-up, none of the patients have presented with any thromboembolic complications or serious bleeding.
Results
The implantable haemodynamic monitor sensor was implanted in 17 Fontan patients during the study period. During a mean follow-up duration of 26 months (range 13–44 months), there were no reports of procedure-related complications, thromboembolism, or device displacement. Of these, 10 patients returned to their residence at a higher altitude after the device implantation (average altitude 5100 feet above sea level, range 2625 feet–6820 feet). Majority of these patients resided in rural communities with limited access to reliable working internet (Table 1). Most families lived remotely (average distance from centre ∼160 miles, range = 100–400 miles), and drove several hours for in-person follow-up or procedures. None of these locations has any paediatric cardiac catheterisation laboratories equipped to implant or calibrate the device. Majority of the patients (15/17, 90%) were below the age of 18 years (range 9–18 years) at the time of implant. The indication for device in these patients included one or more signs of failing Fontan physiology, including worsening Fontan-associated liver disease, recurrent protein-losing enteropathy, and depressed single ventricle systolic function with clinical signs of heart failure. Patients were followed up for a median duration of 26 months after implant (range 13 months–44 months). During follow-up, 2 patients (patient 1 and patient 9) underwent a cardiac transplant. The implantable haemodynamic monitor device was retained and recalibrated for pulsatile circulation during their surveillance cardiac catheterisation, and the device is currently being utilised for post-transplant heart failure management.
Demographic and haemodynamic characteristics of the study patients with indication for IHM implantation
1– Extracardiac Fontan; 2 – Protein losing enteropathy; 3 – Fontan associated liver disease.
Device performance during follow-up, lessons learnt, and interventions performed
Table 2 outlines the overall performance of the sensor in our group. The device readings at home correlated well with the readings obtained at the hospital prior to discharge in all the patients (paired t-test, p value > 0.5). Immediately after device implant, patient 13 was noted to have very low home readings. This was found to be due to a defective home interrogation unit, and the issue was resolved after replacement of the “pillow” by the manufacturer.
A summary of IHM sensor issues and effectiveness of interventions applied for optimising IHM performance on patients with Fontan circulation
# Altitude above 2000 feet above sea level, *paired t-test used to compared mPAP at implant with baseline pre-discharge reading (p > 0.5, not significant).
During follow-up, 7 patients were noted to have sensor reading discrepancies, and several challenges were identified and addressed by our team. Aetiology of these issues remained multifactorial, therefore, some patients required more than one intervention to address these issues. Here is a summary of lessons learned and interventions applied to our study group.
Difficulty in performing home readings: unreliable internet connectivity
Remote monitoring was particularly challenging in patients from remote, rural areas, with at least 5 patients unable to provide consistent readings due to unreliable internet connection (patients 2, 5, 10, 11, and 17). The device representative has been working with these families and has provided information on alternate methods to connect the device, including connection to the cell phone and landlines.
Sensor reading discrepancy due to motion artefact and application of “Fontan specific settings”
Patients 5, 6, 8, 15, and 16 were able to transfer home readings, but the home readings were significantly lower than pre-discharge readings. Manual review of these readings and waveforms was performed systematically with the help of the device engineering team. There appeared to be a measurement error as the device misinterpreted the non-pulsatile waveform with troughs and peaks due to movement artefact as “pulsatile circulation.” As these recordings are done at home, the patient or caregiver does not see any waveforms, and do not typically receive an error notification by the device; therefore, the recordings were transmitted as valid signals. Corrective intervention was applied on patients 6, 8, and 16 by applying “Fontan specific settings” to the device, which included overriding pulsatility, and a minimum of 2 readings were repeated to confirm the effectiveness of the intervention (Figure 2). These patients have had no further issues in their home readings. Patient 16 suffered an acute decompensation in her haemodynamic status since backend recalibration and had died about 2 months after the recalibration. These Fontan settings were also applied to patients 5 and 15, but the patients have had difficulty performing home readings with limited resources and barriers due to their remote residence. Their clinic readings have been consistent and indicate that the sensor is still functional, however, we have experienced difficulty with home monitoring.
IHM sensor discrepancies and effectiveness of interventions applied on Fontan patients.
Sensor reading discrepancy due to upgrade of interrogation unit
Patients 4, 6, and 8 had their device implanted in 2022 and had consistent device readings until March 2023, when their device readings showed a significant drop in the measured mean pulmonary artery pressures. This coincided with the period when the hospital electronics interrogation unit was upgraded to a newer model, which created an offset of mean pulmonary artery pressures, measured by the previous unit at the time of initial calibration. It was discussed that the gold standard for recalibration would be a repeat haemodynamic cardiac catheterisation; however, there were many risks that were associated with this approach. Some of the risks of elective cardiac catheterisation in failing paediatric Fontan patients involve exposure to general anaesthesia and patient and family inconvenience in travelling from remote locations. Therefore, recalibration via confirmatory cardiac catheterisation was deferred until clinically indicated. We then worked with the device engineering team to perform backend recalibration of the sensors. The original calibration values at the time of implant (using the original hospital electronics system) were compared to the most recent values taken in clinic during follow-ups with the new electronics system. The delta between the two readings was utilised to re-baseline the sensor. The device readings have been consistent and close to expected during the follow-up on these patients (Figure 2).
Sensor drift
Patient 7 had accurate readings until about 19 months after implant and then had sudden change in readings, with significantly lower mean pulmonary artery pressures compared to baseline. Further review of the waveforms was thought to be due to a rare sensor drift. The device position was confirmed with a chest X-ray and had not moved since the device placement. Patient was advised to undergo recalibration of the device via haemodynamic cardiac catheterisation; however, the family has finally agreed to undergo a repeat haemodynamic procedure due to her worsening cyanosis, and the device is pending recalibration.
Recalibration of the device during haemodynamic cardiac catheterisation
After implant, 2 patients (1 and 9) underwent heart transplant, and their device was recalibrated. Patient 11 and 17 presented with a failing Fontan physiology, and due to their deteriorating clinical status, they underwent a repeat cardiac catheterisation, when their device was interrogated. For patient 11, device sensor reading was about 2 mmHg lower than the live pulmonary artery pressure and the device was recalibrated. For patient 17, device sensor readings exactly matched the live pulmonary artery pressure readings; therefore, no recalibration was necessary.
Subgroup analysis of the patients with sensor discrepancy
Patients who developed sensor reading issues during follow-up (n = 7) were compared with the 10 patients who continued to have reliable readings. Patient living at high altitude were more commonly in the discrepancy group (6/7 versus 4/10, p = 0.2); however, this was not statistically significant. Presence of stent in the left pulmonary artery or Fontan pathway was also not significantly different in the two groups (2/7 versus 1/10, p = 0.3) calculated.
Discussion
As a large single ventricle centre with over three years of experience implanting implantable haemodynamic monitor sensor in young patients with Fontan physiology, Reference Bhat, Graziano, Garn and Franklin5,Reference Sabati, Chesney, Montchal and Bhat9 we have had a unique opportunity to assess the performance of the sensor when subjected to combined challenges of interpreting non-pulsatile circulation in the setting of higher altitude and remote rural regions of Arizona. Our paper aims to introduce readers to the significant potential of implantable haemodynamic monitor in managing the high-risk Fontan circulation, while also sharing key lessons learned in addressing some of the challenges encountered. To the best of our knowledge, this represents the largest series of Fontan patients evaluating implantable haemodynamic monitor performance in remote, rural, and high-altitude settings. Designed to reduce the need for in-person visits and resource-intensive procedures, the device holds appeal for patients living far from major tertiary care centres. However, our findings underscore that this very population may also present the greatest challenges for the effective implementation of such technology, due to the complex barriers to care they routinely face.
The Fontan circulation depends on the passive blood flow into the pulmonary circulation, and multiple physiological states such as exercise, dehydration, and change in altitude may impact overall resistance in this circuit, and eventually cardiac output. Implantable haemodynamic monitor is, therefore, a very promising tool for monitoring the change in Fontan pressures in response to these physiological stressors. We have previously demonstrated the use of this device in assessing response of the Fontan circulation to exercise. Reference Sabati, Chesney, Montchal and Bhat9 The true extent of higher altitude on Fontan circulation, however, is uncertain. Reference Zhou, Malhotra and Yu10–Reference Malhotra, Ivy and Mitchell16 Our study further illustrates that the true impact of higher altitude may be difficult to separate from the effect of socio-demographic factors, and access to care.
Our experience with device when used in the unique setting of non-pulsatile Fontan circulation highlights a potential risk for misinterpretation of the pulmonary artery waveforms by the sensor. As the sensor is primarily designed and tested on pulsatile pulmonary circulation, small changes in the waveforms, such as undulations related to patient movement or respiration, may cause error in the calculated readings. These errors can be significantly reduced if “Fontan specific settings” are applied at the time of implant, but may not be completely prevented. In case of suspicious readings, we recommend that the clinicians perform careful evaluation of the waveforms before accepting the reading and applying it for clinical decision-making.
There are isolated case reports of device sensor error when exposed to frequent changes in altitude. Reference Hamadeh, Milligan and Minniefield19 Hamadeh and colleagues reported an error in the device recordings in an adult heart failure patient on ventricular assist device who was exposed to frequent changes in altitude. In response to post-marketing reports, the manufacturer performed device testing and acknowledged a possible sensor error when the implantable haemodynamic monitor device is used at an altitude significantly higher than the level at calibration. Reference Boltz20,21 According to this notification, while an increase in measurement inaccuracy was possible at an altitude ≥2,000 feet above sea level, the readings were stable and repeatable (within 3.3 mmHg) when taken under consistent conditions (elevation, atmospheric, etc.), based on device testing. Ideally, device implantation and initial calibration would be performed at the same altitude where the patient permanently resides. However, due to the absence of fully equipped cardiac catheterisation laboratories in rural areas—and the logistical challenges of performing invasive procedures in paediatric single ventricle patients—we were unable to calibrate or recalibrate the device near the homes of these patients. Nevertheless, periodic recalibration may be more feasible in other patient populations, such as adults, and should be pursued whenever possible if sensor malfunction is suspected.
Based on the adult studies, the implantable haemodynamic monitor sensor is very stable with a very low expected “drift rate,” with only about 0.6% requiring recalibration in post-marketing analysis. Reference Vaduganathan, DeFilippis, Fonarow, Butler and Mehra24 Data are limited on long-term performance of these devices in CHD patients. Salavitabar et al reported that recalibration of implantable haemodynamic monitor was required in 6 out of 8 adult palliated CHD patients during their follow-up, with pressure errors noted between 1 and 6 mmHg. Reference Salavitabar, Bradley and Chisolm8 A recent review of this device use in CHD patients also confirmed this higher drift rate. Reference W.H.Marshall, Rajpal and Mah7 None of these patients were exposed to extreme change in altitude after device calibration, as reported in our study. We noted this sensor drift in 1 out of 17 patients (6%), and we plan to recalibrate the device with a haemodynamic cardiac catheterisation.
We did not identify any significant risk factors associated with sensor discrepancies when comparing patients based on age at implant, type of Fontan conduit, degree of altitude change, presence of a fenestration, or other metallic hardware such as a left pulmonary artery stent. The device implant site and technique were consistent across all patients. Most of the discrepancies observed in our study appeared to be related to minor and modifiable factors—such as the need to apply Fontan-specific settings to account for non-pulsatile circulation, attention to potential offsets caused by changes in interrogation units, and manual review of suspicious waveforms. However, once a sensor error is suspected, the gold standard for confirming device accuracy remains invasive haemodynamic assessment via cardiac catheterisation. We found that this can be particularly challenging in paediatric Fontan patients, especially those residing in remote areas. Even though these results are based on a relatively small sample size, these findings are important clinically and should alert the clinician to verify validity of recorded data before it is used to make clinical changes.
Based on our experience, we have modified our practice and become more selective in patients we refer for implantable haemodynamic monitor. We consider patient’s access to technology, reliable internet or cellular connectivity, and willingness to provide consistent, regular home readings, so that discrepancies can be detected early and addressed as soon as possible. Readers and future users may find our experience useful for candidate selection.
Despite the challenges encountered and the interventions required, implantable haemodynamic monitor has proven to be a valuable tool in our patient population. Among patients who adhered to our team’s recommendations for regular home transmissions, we were able to detect early signs of rising Fontan pressures, enabling timely clinical follow-up and escalation of therapy. Home measurements were convenient for most families, and many teenage patients consistently performed readings independently, without parental assistance. Additionally, families from remote rural areas reported feeling empowered by having access to the device, as it offered reassurance and avoided the need for long travel to the nearest hospital when symptoms arose. When implemented in the appropriate clinical context—with engaged patients and families, and close monitoring by a dedicated care team—implantable haemodynamic monitor holds strong potential as a tool to overcome barriers to care for this vulnerable population.
Our study has several limitations. The sample size is relatively small, and results cannot be generalised. While some adult studies have shown the impact of altitude on implantable haemodynamic monitor performance, Reference Hamadeh, Milligan and Minniefield19,Reference Al-Abboud, Al-Saffar and Afzal25 our study is not designed to clearly assess that effect, in the absence of the ability to validate the device readings at varying altitudes. Additionally, the retrospective design of the study does not allow assessment of true impact of the altitude on Fontan pressures, in the absence of confirmatory cardiac catheterisation at the higher altitude on the same individual. Looking ahead, it may be worthwhile to validate device measurements directly by performing haemodynamic cardiac catheterisation on the same patient at two different altitudes. This would provide a more robust assessment of the device’s accuracy and its utility in managing Fontan patients living at varying altitudes. So far, our Fontan team has not been able to perform these haemodynamic measurements due to logistic reasons.
Nevertheless, the study’s strengths should be noted. To our knowledge, this is the largest series of Fontan patients where an implantable haemodynamic monitor has been used to monitor Fontan pressure in remote, high-altitude settings.
In conclusion, our study demonstrates that the implantable haemodynamic monitor sensor is generally reliable in Fontan circulation, supporting its role in guiding clinical decision-making. While implantable haemodynamic monitor remains a promising and valuable tool for non-invasive monitoring of these patients, its full potential will depend on addressing the technical challenges unique to this Fontan physiology. We hope that the lessons we learnt and potential solutions to address these challenges may be helpful to other current and future users of this device, and further improve its role in managing this challenging patient population.
Acknowledgements
The authors would like to acknowledge Fontan clinic nurse coordinator William Chesney, MSN and interventionalist, Byron Garn MD for their help with providing data for the manuscript.
Author contribution
DPB: Conceptualised and designed the study, collected data, analysed data, drafted, edited, and finalised the manuscript. AS: Edited and revised and reviewed the final manuscript. SA: Edited, revised and reviewed the final manuscript. JNG: Edited, revised and reviewed the final manuscript.
Financial support
This research was not funded.
Competing interests
DPB, AS, SA and JNG have no financial disclosures.
