Extracorporeal membrane oxygenation in a low-volume whole nation cardiothoracic centre: short- and long-term mortality, survivors` quality of life and symptoms of post-traumatic stress disorder Authors: Inga Lara Ingvarsdottir, Solveig Helgadottir, Liney Simonardottir, Göran Dellgren, Tomas Gudbjartsson Published: 2025 ABSTRACT AbstractBackgroundFor over three decades ECMO-treatment has been provided in our low-volume nationwide cardiothoracic centre in Iceland. This research aims to examine both the short- and long-term results of all adult ECMO-cases at our centre, including health-related quality of life (HrQoL), return to work status and symptoms of post-traumatic stress disorder (PTSD).MethodsA retrospective study including all adult patients treated with ECMO in a single university cardiothoracic centre in Iceland between 2002 and 2022. Five-year survival was evaluated with Kaplan-Meier estimates, HrQoL with the EQ-5D-5 L instrument and PTSD with the LEC and PLC-5 check lists assessing traumatic life events and PTSD symptoms, respectively. Median follow-up time was 6.5 yrs.ResultsNinety-three patients received ECMO during the study period, an average of 5.2 patients per year since 2006. Sixty-six patients received venoarterial (VA) ECMO, of whom 28 (42%) could be weaned and 21 (32%) survived to discharge. Twenty-seven patients received venovenous (VV) ECMO, with 18 (67%) surviving to discharge. Twenty-four out of 30 living patients completed follow-up questionnaires. No problems with an EQ-5D-5L domain was reported 21 times in the VA ECMO group and 33 times in the VV ECMO group. Three out of four patients reported problems with anxiety or depression. Six patients (25%) met screening criteria for PTSD. At the end of follow-up, 2/12 VA ECMO and 8/12 VV ECMO patients had returned to work .ConclusionsECMO treatment can be provided with acceptable short- and long-term mortality in a low-volume centre. Nonetheless, opportunities for improvement could be identified, particularly regarding VA ECMO in cardiac arrest. At long-term follow-up, problems with anxiety and/or depression were most commonly observed and 25% met screening criteria for PTSD. Most VA ECMO surviors did not return to work.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13019-025-03701-6.Keywords: Venovenous ECMO (VV ECMO), Venoarterial ECMO (VA ECMO), Post-traumatic stress disorder, PTSD, EQ-5D-5L, Health-related quality of life, Return to work status, Mortality, Outcome Introduction Extracorporeal membrane oxygenation (ECMO) for severe and refractory respiratory and/or cardiac failure has been offered in Iceland since 1991, very sporadically for the first years but more commonly in the last two decades [1]. It is offered in a single university cardiothoracic centre in Reykjavik, a low-volume centre serving fewer than 400,000 inhabitants. Given the countries geographical position, referrals of these extremely vulnerable patient groups would be very challenging, which is why we have continued to provide this potentially life saving treatment at our hospital. ECMO-treatment is both complex and resource demanding and therefore it is important to assess the outcomes, particularly in a low-volume centre such as ours. As reflected in reports from the Extracorporeal Life Support Organization (ELSO), reports on the outcome of ECMO have as yet mostly focused on survival to hospital discharge or transfer, rather than long-term survival and quality of life (QoL) [2]. A few previous studies have shown that survivors of ECMO-treatment in the intensive care unit (ICU) can sustain long-term sequele of their life-threatening illness, such as physical problems, depression and anxiety, that reduces their HrQoL [3]. Furthermore, symptoms of post intensive care syndrome (PICS) including cognitive, psychological and physical decline are common after the ICU stay [4, 5]. These symptoms are, however, less well studied in ECMO-patients specifically. We therefore wanted to evaluate the overall outcome of ECMO treatment at our centre, focusing not only on short-term outcome but also on long-term survival and health-related quality of life (HrQoL), including symptoms of post-traumatic stress disorder (PTSD). Materials and methods This was a retrospective study with a follow-up questionnaire sent to all the survivors in October 2022 and 2023. The study was based on a nationwide cohort of all adult patients that received venoarterial (VA) or venovenous (VV) ECMO for all indications at Landspitali University Hospital between 2002 and 2022. The short-term outcome of a part of the VV ECMO patient cohort has previously been partially reported, but they were included in the current study together with new VV ECMO cases, since the main focus was on long-term outcome including HrQoL and PTSD [1]. Approvals for the study were attained from The Icelandic National Bioethics Committee as well as the Landspitali Ethics Committee (appl. number: VSN-21–240-S1). Individual consent was waived for the retrospective part of the study, but obtained from all participants that completed the follow-up questionnaire. Patients and data collection Patients were identified from the centralized ECMO-registry at Landspitali University Hospital, the sole institution offering open heart surgery and ECMO in Iceland. All patients treated with ECMO were included, but those who were converted to VV ECMO from VA ECMO were only included once, as VA ECMO recipients. Data was collected from patient charts. Outcome variables included the following; time on ECMO (days), maximal ECMO flow (L/min), successful weaning, survival to hospital discharge, cerebral performance category (CPC) score at discharge [6] (where a score of 1 or 2 was considered good neurological outcome), one year survival, and complications on ECMO; stroke (ischemic or hemorrhagic, confirmed by radiology), continuous renal replacement therapy (CRRT), bleeding causing hemodynamic instability, withdrawal of anticoagulation or an invasive operation and leg ischemia requiring an operation. Patients were grouped into six different categories according to diagnosis and operative status; (i) Elective or acute cardiac surgery patients were grouped as post-cardiotomy VA ECMO, including if they had a concurrent acute myocardial infarction (AMI); (ii) patients with AMI who were not operated on were grouped as AMI; (iii) A non-ischemic heart failure (NIHF) group including patients with myocarditis and cardiomyopathies; (iv) A Miscellaneous group which included pulmonary embolism, drug intoxication, sepsis, aortic diseases and arrythmias. Other indications were (v) hypothermia and (vi) drowning. CPR prior to ECMO was defined as CPR, with or without return of spontaneous circulation, in the 24 h preceding ECMO installation. Furthermore, ECPR was defined as ECMO installation with ongoing CPR. ECMO installation and patient care We have previously described the decision making, cannulation and subsequent standard of care for the patients on VV ECMO [1]. It has not changed considerably since then, except that after 2018 the dual lumen jugular vein cannula has not been in use. The decision to initiate VA ECMO treatment was made jointly by a cardiothoracic surgeon and anesthesiologist, with a cardiologist often included. Non-surgical patients were cannulated in the femoral vasculature, either percutaneously or with a cut-down technique. Patients who received ECMO for post-cardiotomy cardiogenic shock were often initially cannulated centrally, but the cannulas later moved peripherally if ECMO support was still needed after chest closure. Four out of 31 extracorporeal cardiopulmonary resuscitation (ECPR) patients had ongoing cardiopulmonary resuscitation (CPR) until cardiac bypass was established, but for the other 27 ECPR patients conventional VA ECMO insertion was used. Anticoagulation was traditionally achieved with unfractionated heparin, both during cannulation as well as during the ECMO support. Cardiac recovery was evaluated at least daily by transthoracic (TTE) or transesophageal (TEE) ultrasound, and often pulmonary artery catheter readings. When cardiac recovery was deemed sufficient, a weaning trial was attempted and the patient subsequently decannulated if hemodynamically stable. If lung function required further support, patients were converted to VV ECMO. After discharge from Landspitali hospital, further rehabilitation was planned according to patient needs. Cardiac surgery patients were routinely offered heart-and-lung rehabilitation with physical therapy. After 2017, selected patients were offered to schedule a follow-up visit at the Intensive Care Unit two to three months after their discharge. They were subsequently referred to their primary care or to seek other specialized healthcare if needed. Follow up Survivors were identified in a centralized nationwide register, with last follow-up in May 1 st 2024. Two foreign residents were lost to long-term follow-up. Three different paper surveys were sent by mail to all survivors. This included the EQ-5D-5 L instrument by the Euroquol research foundation (with permission, ID-45323) which consists of both a descriptive system; wherein a patient is asked to state their health by choosing a statement of no [1], slight [2], moderate [3], severe [4] or extreme [5] problems in the following five domains; mobility, self-care, usual activities, pain/discomfort and anxiety/depression, as well as a visual analoge scale (EQ VAS), where the patient is asked to rate his/her/their health on a scale of 0–100 [7]. The Swedish value set was used to calculate the VAS-index value from the five health domains, that was used for comparison, as an Icelandic value set is not available [8]. Symptoms of PTSD following hospital stay were assessed using the PTSD Checklist for DSM-5 (PCL-5), with reference to the ECMO treatment, as well as the Life Event Checklist for PTSD [9, 10]. Answers were assessed for screening criteria for PTSD both by the DSM-V critera as well as by the PCL-5 total score [11]. Finally a set of five questions were included, asking of return to work or study, problems with cannulation sites, a subsequent history of deep vein thrombosis or pulmonary emboli, together with any subsequent hospital admissions for heart or lung diseases. Statistics Statistical analysis was performed in R version 4.2.2 (R foundation for Statistical Computing, Vienna, Austria (R 4.2.2.) with RStudio (version 2022.12.0 + 353). Continuous variables were reported as median with interquartile range. The pH prior to ECMO was missing for 9 VA ECMO patients, which were therefore not included in the reported median. Results of the EQ- 5D-5L analysis were compared between the VA and VV ECMO groups using the Wilcoxon test and the Pearsons chi squared test. A p-value of < 0.05 was considered as statistically significant. Five-year survival was analysed using Kaplan-Meier estimates. Results There were 93 adult patients that received ECMO for cardiogenic and/or respiratory failure between September 2002 and December 31 st, 2022; 66 (71%) with VA ECMO and 27 patients (29%) with VV ECMO. The overall survival for all patients to discharge was 42%; 32% and 67% for VA and VVECMO, respectively (Fig. 1). Table 1 summarizes patient demographics for all patients together and Fig. 2 demonstrates their 5-year survival estimates according to Kaplan Meier. The patient cohort and follow-up Table 1Patient demographicsCharacteristicVA ECMO, (n=66)VV ECMO, (n=27) Female sex (%) 23 (35)12 (44)Age in years, Median (IQR)53 (43–67)34 (24–50) Hypertension (%) 15 (23)8 (31) Chronic pain (%) 3 (5)3 (11) Psychiatric disease (%) 12 (18)9 (33) Diabetes mellitus (%) 9 (14)1 (3.7) Previous cardiac surgery (%) 6 (9.1)1 (3.7) Smoking (%) 12 (18)3 (11) Cardiovascular disease (%) 21 (32)2 (7.4) Chronic renal failure (%) 2 (3.0)0 (0)pH prior to ECMO, Median (IQR)7.08 (6.9–7.26)7.29 (7.19–7.40)Pre-ECMO days on ventilator, Median (IQR)0.1 (0.1–0.5)1 (0.5–7.5) Murray score (IQR) n/a3.7 (3.5–3.75) Acute respiratory distress syndrome (%) 12 (18)26 (96) Cardiopulmonary resuscitation (%) 37 (56)0 (0) Extracorporeal cardiopulmonary resuscitation (%) 31 (47)0 (0) Acute myocardial infarction (%) 26 (39)0 (0) Patient demographics Kaplan Meier survival probability for patients that underwent VA- and VV ECMO VA ECMO characteristics and short-term outcome Of the 66 patients that received VA ECMO, 35% were female and the median age was 53 yrs. The indications for VA ECMO are summarized in Fig. 3a. All the VA ECMO patients had refractory cardiogenic shock and in five cases there was a concomitant severe respiratory failure present. Thirty patients had ECMO following cardiac surgery, of which 21 procedures were acute. Twenty-two of the 30 patients suffering post-cardiotomy shock received a central aortic cannulation for their ECMO. Thirty-seven patients had cardiac arrest prior to ECMO; of which 27 had ongoing chest-compressions during initiation of VA ECMO. Four patients had ongoing chest compressions untill the start of cardiopulmonary bypass (CPB), followed by VA ECMO post-operatively, as weaning from CPB was not possible. Median pH prior to ECMO or bypass was 7.08 (IQR 6.9–7.26). For ECPR patients, median time-to-ECMO was 90 min (IQR 68–120 min.), but in 9 patients the time-to-ECMO was missing. VA ECMO cases by etiology (a) and year (b), stratified by survival to discharge. AMI = Acute myocardial infarction, NICM = Non ischemic heart failure, ECPR = Extracorporeal cardiopulmonary resuscitation Short-term outcomes for VA ECMO are summarized in Table 2; Fig. 3. Thirty-four patients could not be weaned off ECMO nor bridged. Three patients were converted to VV ECMO, and one patient received a left ventricular assist (LVAD) device after a transfer to a transplantation centre overseas, but died shortly thereafter. Altogether 21 patients (32%) survived to hospital discharge, 18 with a good neurological outcome at discharge. For the subgroup of ECPR; 6 out of 31 patients survived (19%). The most common complications were the need for continuous renal replacement therapy (42%) due to acute kidney injury (AKI) and severe bleeding (36%) that necessitated acute re-operation, discontinuation of anticoagulation and/or massive transfusions. Table 2VA and VV ECMO outcomeCharacteristicVA ECMO (n=66)VV ECMO (n=27) Cannulation Peripheral (%)44 (67)20 (74) Central (%)22 (33)n/a Double-lumen (%)n/a7 (26) Days on ECMO, Median (IQR) 4 (2–7)9 (5–21) Max circuit flow (L/min), Median (IQR) 3.85 (3.28–4.50)4.20 (3.90–4.90) Leg ischemia (%) 9 (14)0 (0) Bleeding (%) 24 (36)7 (26) Re-operation (%) 16 (24)1 (3.7) Stroke (%) 3 (4.5)0 (0) Continuous renal replacement therapy (%) 28 (42)8 (30) Intra-aortic balloon pump (%) 34 (52)n/a Weaning No (%)34 (52)7 (26) Yes (%)28 (42)20 (74) To Venovenous ECMO (%)3 (4.5)n/a To a Left ventricle assist device (%)1 (1.5)n/a Survival to discharge (%) 21 (32)18 (67) Cause of death (%) Cardiac death (%)24 (51)0 (0) Anoxic brain injury (%)8 (17)0 (0) Multiple organ dysfunction (%)8 (17)4 (50) Other (%)7 (15)0 (0) Respiratory failure (%)0 (0)4 (50) CPC score at discharge (%) 1–2 (%)18 (27)17 (63) 3–4 (%)3 (4.5)1 (3.7) 1 year mortality (%) 45 (68)9 (33) VA and VV ECMO outcome VA ECMO long-term outcome All patients that survived to discharge were alive at one year follow-up, but three patients died after that. The follow-up questionnaire was sent to 16 patients, of whom 12 responded (Fig. 1). Of the responders, 7 were male, the median age at follow up was 60 yrs (range 44–77 yrs.) and with a median of eight years since ECMO treatment (range 1–15 yrs). Two out of 12 (17%) had returned to work or school, 6 months after the hospitalization (8%) and 10 years after (8%), respectively. Four patients reported problems at the femoral cannulation site; including one patient with decreased motor function below the knee, necessitating a walking aid for mobilisation. Answers to the EQ-5D-5 L instrument are summarized in Fig. 4a. Least problems were reported in the self-care domain but the largest impact was found with regards to anxiety or depression, where 10 patients (83%) reported any problems, and 5 patients (42%) reported moderate or severe problems. Regarding overall state of health, the median VAS score was 50 (IQR 45–75) and the median VAS-index score was 63 (IQR 48–71) (Figs. 4c and d). EQ-5D-5 L outcomes (a) EQ-5D-5 L domains for VA ECMO patients (b) EQ-5D-5 L domains for VV ECMO patients (c) Box plot describing VAS score, (d) Box plot describing VAS index score VV ECMO characteristics and short-term outcome During the study period 27 adult patients received primary VV ECMO; thereof 26 for respiratory failure due to acute respiratory distress syndrome (ARDS) (Supplementary Table S1). The median age was 34 yrs (range 18–74 yrs) and 56% were males. The median time on ventilator prior to VV ECMO was one day (range 5 h – 18 days), the median pH prior to ECMO was 7.29 (IQR 0.21) and the median pO2/FiO2 ratio 62 mmHg (range 30–118). Altogether 20 patients (74%) were weaned off VV ECMO and 18 (67%) survived to discharge. All three patients with H1N1 infection and the two patients with COVID-19 survived to discharge. Two survivors were foreign residents and lost to follow-up. VV ECMO long-term outcome All patients that survived to discharge were alive at one year follow-up, however, two patients died later on. The follow-up questionnaire was sent to 14 VV ECMO patients and 12 of those responded (Fig. 1). Five out of the 12 were males, the median age at follow up was 44 yrs. (range 22–57 yrs.) and with a median of four years since ECMO (range 1–21 yrs.). Eight patients (67%) had returned to work or school, earliest 4 months after discharge and latest 8 years after discharge. Two patients reported unspecific problems at the femoral cannulation site, and one patient had suffered a deep vein thrombosis in the leg. Answers to the EQ-5D-5 L instrument are summarized in Figs. 4b. VV ECMO patients reported that they had no problem in any domain 33 times in total. Similar to the VA ECMO patients, the VV ECMO patients had the least problems in the self-care domain, but were most commonly affected in the anxiety and/or depression domain, where 67% of the patients reported any problems and five patients (42%) had moderate to extreme problems. The median VAS-score and VAS index score was 78 (IQR 65–80) and 68 (IQR 58–77), respectively (Figs. 4c and d). Symptoms of post-traumatic stress disorder Six patients in total (25%), three in the VA and three in the VV ECMO group, met the screening criteria for PTSD. Of those, five (83%) reported another, unrelated, extraordinarily stressful life event according to the life event checklist. Discussion This study showed that patients that underwent VA ECMO in a low volume centre, most often due to cardiogenic shock related to MI or cardiac surgery, had a 32% survival to hospital discharge. All survivors remained alive at one year follow up, and only three deaths had occured at a later follow-up. Furthermore, 67% of VV ECMO patients survived to discharge, all of whom were alive at one-year follow up. This can be considered a good outcome of VV ECMO treatment, compared to the 63% survival to dishcarge or transfer that is reported by the ELSO-registry which includes over 130.000 cases worldwide [2]. The survival reported here is also in line with results from randomized trials on VV ECMO reporting 63–65% survival [12, 13]. The survival figures reported here for VA ECMO are, however, lower than the 44% survival to discharge or transfer found in the ELSO-registry [2]. This difference could be attributed to low volumes in Iceland, with a current population of just under 400.000 inhabitants. Indeed, ELSO recommends a minimal annual case load of 6 ECMO-cases for cost-effectiveness and to maintain centre competence, which our single nationwide centre has reached only a few years during the 22 year study period [14]. Importantly, however, the survival rates are comparable to multiple other studies that have reported survial for post-cardiotomy VA ECMO in the 29–41% range [15–19]. Furthermore, since Landspitali is not a heart- and lung transplantation centre, our patient-cohort does not include VA ECMO patients with primary cardiac graft dysfunction, that generally have a more favorable prognosis compared to other VA ECMO patients [15, 16, 20]. Nevertheless, we observed less favorable survival for the AMI cases, where only two out of 13 patients survived (15%), as well as for the ECPR cases, with only 6 out of 31 patients surviving (19%). Notably 85% of the AMI VA ECMO cases underwent ECPR. Additionally, 27% of the post-cardiotomy patients underwent ECPR, most often leading up to an acute coronary artery bypass graft operation; necessitating VA ECMO to wean off from cardiopulmonary bypass. ECPR has been a part of our care for more than 20 years, but in Iceland we observed a spike in its use between 2017 and 2018, with only one out of 8 patients surviving. The median CPR time of 90 min until ECMO/CPB was established was also striking, as longer time-to-ECMO has been associated with poorer outcome [21]. There can be several reasons for this long CPR time. One is that for both hypothermic CA as well as CA subsequent to drowning; the retrieval of patients, including transportation by helicopter or ambulance, can be a time-consuming process. It could also partly be explained by a lack of specific ECPR protocols and training at our centre in the earlier time-periods of the study. This evolution could be viewed as a part of a paradigm shift in the treatment of cardiac arrest. Often it is easier to adopt new treatment possibilities at a small center, which was the case with ECPR at Landspitali. However, as both the experience as well as international publications grew, ECPR guidelines were developed in 2018. They were to a large extent based on similar guidelines from the Sahlgrenska University Hospital in Gothenburg, and included variables such as no-flow or low-flow time, and the presence of a shockable rhythm [22]. However, an ECPR bundle of care for cardiac arrest, stretching from the pre-hospital environment to an ECMO decision and initiation of treatment, was still missing. Subsequently, after 2018, the use of ECPR subsided. Patient selection is a key factor in ECMO-treatment and it is important for all centres to constantly evaluate the outcomes of ECMO-use and adapt the treatment. Therefore, until there is a better system for ECPR, especially for OHCA, a more restrictive approach has been adopted in Iceland, adhering to patient selection guidelines. When it comes to long-term results, increased awareness of the quality of life after ECMO treatment can be helpful to improve patient outcome. Anxiety and/or depression might be largely influencing HrQoL in the survivors, as 75% of responders reported problems in the anxiety and/or depression domain, and 42% of the patients reported moderate to extreme problems. These figures are significantly higher than reported in the Icelandic population norm (7.7%) [23], but comparable to studies on ICU survivors in general. There the incidence of clinically significant depression and/or anxiety seems to lie in the 20–57% range; negatively impacting the HrQoL [24–25]. Incidence of significant PTSD symptoms was 25% in our study, which is comparable to other studies for both ECMO and ICU survivors [24–27]. Numerous other studies have shown that severity of illness is not a predictor of PTSD symptoms, however, prior psychopathology is [25], which also is in line with our findings. These results might encourage ECMO-centres to pay attention to the psychiatric wellbeing of ECMO-survivors; with main focus on those patients risking long-standing psychiatric illness that negatively affect their HrQoL following ECMO-treatment and ICU stay. There were some indicators of a higher HrQoL for VV compared to VA ECMO patients in the currrent study, such as a higher rate of no problems in any domain, as well as a higher return to work status in the VV ECMO group. Social participation has indeed been reported to have an influence on HrQoL [28, 29]. There is limited research available comparing the ECMO modalities, with one review article describing more favorable outcome for VV ECMO patients [3]; but the contrary also has been reported in a small patient cohort [30]. The overall VAS index score for both groups was 66 which is comparable to both ICU survivors as well to many chronic medical conditions [31, 32]. For comparison, a study from a large sample of the Swedish general population showed a VAS index value ranging from 58.2 for those with COPD to 72.8 for those with hypertension [33]. A lower VAS score was also found for those who had a self-reported condition of dejection and anxiety/worry, with a mean VAS index value of 64.6 and 65.8, respectively [33]. Limitations and strength This study has limitations due to its retrospective and single-centre design; including potentially missing data on indications and complications. The study period covered 21 years and during this time there have been advances in both the surgical techniques, ECMO equipment and ICU care. We did not have information on patients where ECMO was considered but not applied for various reasons, and these patients are therefore not included in the study. The variables time-to-ECMO for ECPR and the pre-ECMO lactate values were too often missing to be reliably reported, and therefore not included. Strengths include our long follow-up time with 95% survival information; that is made possible using centralized registries in Iceland, and 80% answering the follow-up questionnaires. Importantly, we can make no assumptions as to the reason some survivors did not respond to the questionnaires. Finally, we cannot assume a causal relationship between the ECMO treatment itself and the patients HrQoL and PTSD symptoms. Conclusion This whole nation study shows that ECMO treatment can be provided with acceptable outcome in a low-volume centre, especially VV ECMO. Nonetheless, opportunities for improvement could be identified, particularly regarding VA ECMO in cardiac arrest. At long-term follow-up, problems with anxiety and/or depression were most commonly observed and 25% met screening criteria for PTSD. Finally most VA ECMO surviors did not return to work. Supplementary Information Supplementary Material 1 Acknowledgements We want to thank all the dedicated personnel at Landspitali University Hospital for their continous contribution to improve patient care. Author contributions ILI, TG and SH were involved in the conception and design of the work, IL gathered and analysed the data and IL, TG, SH, LS and GD drafted or substantively revised the work. Funding This study was funded by the Helga Jonsdottir and Sigurlidi Kristjansson memorial fund and Landspitali Science Fund. Data availability The data underlying this article will be shared on reasonable request to the corresponding author. Declarations Ethics approval and consent to participate Approval for the study was attained from the The Icelandic National Bioethics Committee as well as the Landspitali Ethics Committee (appl. number: VSN-21-240-S1). Individual consent was waived for the retrospective part of the study, but obtained from all participants that completed the follow-up questionnaire. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data underlying this article will be shared on reasonable request to the corresponding author.