A Clinical Audit of Cardiovascular Risk Factors and Disease in Patients with Rheumatoid Arthritis - SURF-RA
Full Text
INTRODUCTION
The increased risk of cardiovascular disease (CVD) in patients with rheumatoid arthritis (RA) has been known for several decades and there has been a high focus on CVD prevention in this patient group. The elevated CVD risk in RA is only partly attributed to a high prevalence of traditional risk factors.1 RA is a chronic systemic inflammatory disease, and the chronic systemic inflammation is an independent CVD risk factor.2,3 CVD as a co-morbidity of RA is often overlooked and undertreated, and also management of CVD risk factors may be frequently neglected.4,5
Clinical audits can be applied to monitor data recording and management, measure clinical performance against guideline standards, and inform both appropriate treatments and the modification of recommendations to improve quality of care in routine practice.6 Robust data from audits can also be used by individual health professionals to improve their practice in response to information about their performance.7 Clinical audits are thus essential tools to monitor guideline implementation in clinical practice and to facilitate improved clinical performance.8 The SUrvey of cardiovascular disease Risk Factors (SURF) is a large and well-established audit of CVD risk factor management which has been performed in patients from primary care and in patients with coronary heart disease (CHD),9,10 and which currently is being performed in patients with stroke, chronic obstructive pulmonary disease, systemic lupus erythematosus and antiphospholipid syndrome.
The SURF in patients with RA (SURF-RA) is a new contribution to the SURF audit family in which CVD and its risk factors, RA treatment and use of CVD preventive medication are evaluated in RA patients. Since the risk of atherosclerotic CVD (ASCVD) and CVD death is increased in patients with RA, the overarching goal of SURF-RA is to improve CVD prevention in this vulnerable patient group.
METHODS
Data from established cohorts and on consecutively examined patients in the time period 2014 to 2019 were included. The participating centres were divided into the following geographical areas: Western Europe, Central and Eastern Europe, Mexico, North America (USA and Canada), and Asia (Supplementary Table S1). The SURF-RA audit was performed to allow for quality improvement of CVD risk management in RA patients. It conforms to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the Data Protection Officer (DPO) at Oslo University Hospital (2017/7243) and a general data protection regulation (GDPR) evaluation was performed by the DPO at Diakonhjemmet Hospital (10/10-2018). Due to pseudonymization of data, a Data Protection Impact Assessment (DPIA) was deemed not necessary. Due to the quality assurance format, informed patient consent from the patient was not collected. Specifications on data protection, security and ethical aspects are described in Supplementary Data 1.
Supplementary Table S1. Number of patients by centre, country, and geographic region.
Data collection
Data were collected on a one-page audit form which also included definitions of terms and instructions for clinicians. From the participating public and private outpatient clinics of rheumatology and cardiology, patients with RA aged >18 years were included. Apart from the age criterion, there were no exclusion criteria.
Demographic data included year of birth and sex. The following RA disease-related variables were recorded: Rheumatoid factor (RF) and anti–citrullinated protein antibody (ACPA) positivity, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), use of anti-rheumatic medication, as well as Disease activity score using 28 joints with either CRP or ESR (DAS28CRP and DAS28ESR).
The presence of objectively confirmed ASCVD was noted. Registered CVD risk factors were: smoking status (never/previous/current), physical activity (moderate meaning walking or equivalent 30 minutes 3-5 times/week, less or more than this), hypertension, hyperlipidaemia, obesity, diabetes mellitus, and the most recent CVD risk factor measurements. Lipids, glucose and glycated haemoglobin A1c (HbA1c) were recorded if there were measurements within 1 year of inclusion to the study. Moreover, use of lipid lowering agents and antihypertensive treatment was recorded.
CRP, ESR and lipid values were analysed according to each centre’s laboratory standards. For general CVD risk screening, fasting status has been shown not to influence the prognostic value of the blood sample11 and we therefore included both fasting and non-fasting lipid values.
The Cardiovascular Health Index Score (CHIS) was defined by control of six risk factors: blood pressure (BP) <140/90 mmHg or < 140/80 mmHg for patients with diabetes mellitus, low density lipoprotein cholesterol (LDL-c) 2, moderate/vigorous physical activity. The number of risk factors at recommended targets was summed, ranging from 0 to 6. CHIS categories were defined as follows: poor if number of risk factors at recommended target/level was <3, intermediate: 3-4, good: 5-6.10
The presence and level of CVD risk factors were analysed separately for patients with and without established ASCVD (CHD, stroke, and/or peripheral arterial disease [PAD]).
Risk of future CVD
To compare the risk of CVD across regions, we estimated the 10-year risk of a fatal CVD event by using the European Society of Cardiology’s Systematic Coronary Risk Evaluation (SCORE). Patients were then classified according to prevailing European Society of Cardiology Guidelines at the time of the data collection,2 into the following risk groups: Very high CVD risk: 1) Established CVD, and/or 2) estimated risk by SCORE >10%. High CVD risk group: 1) Presence of diabetes mellitus, and/or 2) TC >8.1 mmol/L, and/or 3) estimated risk by SCORE >5% and <10%, and/or 4) patients treated with lipid lowering medication. Moderate and low risk: estimated risk by SCORE of 1-4% and <1%, respectively.
Statistics
The Kolmogorov-Smirnov test was used to evaluate the distribution of each parameter. Continuous variables are presented as mean with standard deviations (SD) or median with inter-quartile ranges (IQR) as appropriate. Categorical variables are presented as percentages. Between-groups comparisons of continuous variables were performed using the Kruskal-Wallis test, while the Chi-squared test was used for categorical outcomes. No imputation for missing data was done, and each variable was summarized using all reported data.
RESULTS
Patient demographics
The 14 503 patients included were from 53 centres across 19 countries in 3 world continents (Supplementary Table S1). The ethnicity of the participants in the various continent cohorts are also described in Supplementary Table S2. Since age, sex and disease duration were comparable across the different world regions and in the established a consecutively examined patient cohorts (Supplementary Table S3a and S3b), and time span for data collection was only 5.7 years, we merged the data from the established and consecutively examined patients into one cohort, which is used in the following evaluations. In the total cohort, the mean (SD) age was 59.8 (+13.6) years and there was a strong female preponderance (74.5%) (Table 1).
Supplementary Table S2. Ethnicity and RA disease-related factors across regions.
Supplementary Table S3a. Demographic and RA disease specific data on patients from established cohorts.
Supplementary Table S3b. Demographic and RA disease-specific data on consecutively examined patients.
Table 1. Cardiovascular disease risk factors across world regions.
Rheumatoid arthritis disease specific variables
More than half of the patients were RF and/or ACPA positive. Disease duration was quite similar across the geographic regions, the mean varying 9.9-12.6 years (Supplementary Table S2). RA disease activity evaluated by composite measures was on average low, and this was stable with increasing age (Figure 1A). Inflammatory markers and DAS28-ESR were highest in Asian RA patients (Supplementary Table S2).
Cardiovascular disease
In all patients, any CVD was present in 17% and ASCVD in 15%. The most common ASCVD was CHD and PAD (7-8%), while stroke, heart failure and atrial fibrillation were present in 3-4% (Figure 2A). ASCVD showed considerable geographic variation, being very low in India (2%) and Mexico (3%) but nearly 40% in Russia (Figure 2B).
Cardiovascular disease risk factors
CVD risk factor levels are reported in Table 1. Mean BMI was 27.4 kg/m2. The highest mean BMI was recorded in North America, while the lowest BMI values were found in the Asian cohort (mean 29.7 kg/m2 and 23.8 kg/m2, respectively). However, the mean BMI in the total cohort was >25 kg/m2, which is classified as overweight. In other words, less than 50% were of normal weight and this was stable with increasing age (Figure 1B).
Mean TC was 5.0 mmol/L, while mean LDL-c ranged from 2.5 mmol/L in North America to approximately 3.0 mmol/L in Central and Eastern Europe. In the whole cohort, few had hypercholesterolemia (< 1%) with TC>8.1 mmol/L. Although the presence of an LDL-c level > 2.5 mmol/L varied across the world regions (46% in Mexico and 73% in Central and Eastern Europe), after 50 years of age the proportion of patients with LDL-c > 2.5 mmol/L decreased (Figure 1C).
BP was on average in the normal range, 128/77 mmHg. Nevertheless, hypertension (BP >140/90 mmHg, and/or self-reported hypertension, and/or use of antihypertensive medication) was present in 2/3 of the patients, this proportion being highest in Central and Eastern Europe (75%) and lowest in Mexico and Asia (36% and 49%, respectively). Overall, BP >140/90 mmHg became more common with age (Figure 1D). Of patients not using antihypertensive medication, an average of 17% had BP > 140/90 mmHg, varying from 13% in Mexico to 25% in Central and Eastern Europe.
Current smoking was most common in Central and Eastern Europe (29%), and least common in Asia, Mexico and North America (8-10%). However, North America had a higher proportion of previous smokers compared to the other regions (34%). The number of never smokers was highest in Mexico and Asia (78% and 85%, respectively). There was a decline in current smokers with increasing age (Figure 1E).
On average, a high proportion of patients reported moderate or more than moderate physical activity (58%). In Mexico and North America, however, the majority reported less than moderate physical activity (70% and 65% respectively).
The prevalence of diabetes mellitus was on average 13% and comparable across the geographical areas. This comprised mainly type II diabetes (12%). Although the number of patients with HbA1c above recommended levels was low, it increased slightly with increasing age (Figure 1F).
The distribution of patients in the different CVD risk categories by SCORE was significantly different across the geographic regions (Table 1). Central and Eastern Europe and North America had the worst CVD risk profile according to SCORE (very high and high CVD risk at 58% and 65%, respectively). In comparison, Mexico and Asia had the most favourable CVD risk profile (36% and 29% having very high and high CV risk, respectively). The number of risk factors at target as evaluated by CHIS was also comparable across the geographic regions, regarding poor and intermediate CHIS. Notably, only ¼ had > 4 risk factors at recommended level/target, except in Asia were half of the patients had good CHIS (Figure 3).
Patients with ASCVD were on average older and more often male than patients without ASCVD (Supplementary Table 4). Evaluating the CVD risk factors in patients with and without ASCVD revealed that patients with ASCVD had somewhat lower lipid levels, more hypertension and diabetes, and were less often physically active compared to those without ASCVD (Supplementary Table 4).
Supplementary Table 4. Risk factors in patients with and without atherosclerotic cardiovascular disease.
MEDICATION
Cardiovascular disease preventive medication
Approximately a quarter of all RA patients were on statin therapy, except from the Asian and Mexican cohorts (6% and 16%, respectively). Use of other lipid-lowering agents than statins was infrequent in all continents (1-7%) (Table 2).
The use of any antihypertensive treatment also differed substantially between the cohorts, with a mean prevalence of 29%, and lowest use reported in Western Europe and Mexico (Table 2).
Use of nitrates for angina pectoris was higher in North America (5%) compared to almost no use in Mexico and Asia.
Use of any antidiabetic medication was comparable across the geographic regions, on average 3% using insulin and 8% using any oral antidiabetic agent, although the latter was somewhat higher in Mexico (11% using an oral antidiabetic agent) (Table 2).
Table 2. Cardiovascular medication across world regions.
Anti-rheumatic medication
The number of patients receiving methotrexate was high, especially in Central and Eastern Europe and Mexico (73% and 80% respectively) (Supplementary Table S2). Treatment with biologic DMARDs was most frequent in Western Europe (42%), but these agents were rarely used in Mexico and Asia (5% and 6%, respectively). The opposite was observed for prednisolone, which was used by 48% in Central and Eastern Europe and 63% in Mexico, while 35-38% used glucocorticoids in Western Europe, North America, and Asia. Usage of non-steroidal anti-inflammatory drugs (NSAIDS) ranged from 15-28% in Western Europe and Mexico but was much more commonly used in Central and Eastern Europe, North America, and Asia (56-63%) (Supplementary Table S2).
Missing data
Missing data on premature CVD, lipids, BP, physical activity, and diabetes was most common in North America and Western Europe, although a lack of HbA1c recording was common in all geographic regions (Supplementary Table S5).
Supplementary Table S5. Number of patients missing recorded data and as percentage of the total population.
DISCUSSION
CVD is a major comorbidity in patients with RA, and in this large international audit, we describe the state of CVD risk factors and their management in patients with RA across three continents. Our main finding is that CVD risk factors are still highly prevalent in this patient population. Nearly half of the RA patients were categorized into high or very high CVD risk classes according to the SCORE algorithm, and established CVD was present in one sixth of patients. We demonstrated that room for improvement exists for control of CVD risk factors for many, since three out of four patients had poor or intermediate CHIS scores (≤4/6 CVD risk factors at target level). The SURF audit structure proved feasible in patients with RA in the setting of rheumatology and cardiology outpatient clinics, comparably to that in patients with CHD.9,10
Cholesterol levels are one of the key CVD risk factors, and should be evaluated among RA patients at a minimum of 5-year intervals and if major changes to antirheumatic treatment occur.12 The desired lipid targets in RA are the same as in the general population. Presumably related to systemic inflammatory response, RA patients commonly have low cholesterol levels.13-15 As in the general population, some patients with RA have an increased risk of CVD even at low levels of cholesterol.16 Nevertheless, in the present study, 63% of RA patients had an LDL-c> 2.5 mmol/L, which is substantially higher than what has been described in other RA patient cohorts,17 and more recent surveys from the general population (34-37%).9,18 This may suggest that management of hypercholesterolemia among RA patients is inadequate, especially when considering that nearly half of the patients in SURF-RA were categorized high or very high risk classes according to SCORE algorithm.
Tobacco smoking is a major risk factor for CVD.19 While smoking prevalence is declining across European countries, 17% of RA patients in this survey are still current smokers. Furthermore, current smoking in Central and Eastern Europe was 1.5 to 3 times more prevalent than in the other regions. Smoking may be of exceptional harm to RA patients: not only are they especially susceptible to CVD events,20,21 smoking may also worsen RA-related outcomes.20,22
Previous studies have reported that approximately 60% of RA patients are either overweight or obese.23 Approximately 25% of patients were characterised as obese (BMI ≥30 kg/m2) in our survey, which is lower than reported for patients with CHD in both Eurospire V18 and SURF-CHD,9 but comparable to that described for the general population in Europe, where the average age-standardised BMI in 2017 was 25.8-26.8 kg/m2.24 While being overweight or obese are risk factors for CVD in the general population, both harmful25 and cardio-protective26 effects of overweight/obesity have been shown in patients with RA.
Diabetes was present in 13% of the SURF-RA cohort, which is in line with that reported in RA patients from one large US-based study,27 but twice as high as that reported in another RA cohort from the US.28 Diabetes is linked to overweight and obesity, but it may also be a side-effect of glucocorticoid use. Due to the effectiveness of newer anti-rheumatic medications and local joint injections, the use of glucocorticoids in patients with RA is declining, but was still used by nearly 40% in this survey.
In addition to traditional CVD risk factors, inflammation drives the RA-related CVD risk, and effective suppression of disease activity is key to lower CVD risk among RA patients.12 Treatment with biologic DMARDs was most frequent in Western Europe and rarely used in Mexico and Asia, which may be related to differences in health care resources. Concordantly, Mexican and Asian RA patients had slightly higher DAS28, ESR and CRP compared to other world regions. The question of whether biologic DMARDs have positive or negative effects on CVD risk in RA is debated.29-31 It has been reported that BP is adversely affected by systemic inflammation and there is accumulating evidence suggesting that hypertension is more common in RA patients compared to the general population.32-35 Treat-to-target and tight control of RA disease activity has been effectively implemented, as novel anti-rheumatic medications have been developed in the last decades. These innovative RA disease monitoring strategies may explain the quite low overall disease activity levels across the 3 world continents in our survey. The reported prevalence of hypertension in our patients was high (62%) and possibly related to the systemic inflammation in RA and moderately increased CRP levels (mean 2.9 mg/dL). Although this level of CRP has been shown to increase BP34 and promote CVD risk in the general population,36 it reflects a modest degree of inflammation for RA patients.
In this large international audit, we revealed that the prevalence of ASCVD was particularly high in regions where the presence of CVD risk factors was also high. In Central and Eastern Europe and North America, the presence of ASCVD was 21% and 16%, respectively. Not surprisingly, the percentages of patients having high and very high risk of CVD according to SCORE algorithm were also highest in these two regions, mirroring the high presence of CVD risk factors. Percentages of patients at very high or high CVD risk were 65% in North Americans, 58% in Central and Eastern Europeans, and 39% in Western Europeans, as compared to 29% in Asians and 36% in Mexicans. These differences are explained by Asian and Mexican RA patients being less frequent smokers, having lower BP and TC, and being less obese compared to North American and Central and Eastern European RA patients. Overall, only ¼ had more than 4 CVD risk factors at recommended levels, reflecting inadequate risk factor control.
The limitations of a survey such as SURF-RA should be noted. The variations of the recorded prevalence of CVD risk factors across the 3 continents may reflect the various settings for CVD risk factor recording. For example, the risk factors reported from North America were mostly extracted from primary care patient records, while SURF-RA centres in Western Europe were mostly hospital rheumatology outpatient clinics, and several cardiology outpatient clinics. The centres participating in the audit were either invited through participation of the ATACC-RA network (www.atacc-ra.com) or invited through conference contacts. Therefore, the representativeness of the cohorts in relation to nations or geographic regions is not complete. It is desirable that a data sampling frame to be as representative as possible, but since this was a clinical audit without funding, this was not possible and is a limitation to the project. Also, the included number of patients from the various regions varied. Mexico recruited few patients in comparison with the other regions and data may therefore not be representative for RA patients in this region. On the other hand, a strength of the study is that the three countries that recorded the highest number of patients were geographically spread (USA: n=3226, Greece: n=3286, Norway: n=3544). Audits are not epidemiologic studies and lack control groups, and thus a limitation to this audit is the lack of a non-RA control group. Another limitation is the relatively long inclusion period (2014-2019) in which CVD risk and disease prevalence in RA populations could fluctuate, however these data are of interest as data from comparable international cohorts in this time period is lacking.
Some of the data were extracted from pre-existing registries, which may explain the differences in missing data across the geographic regions. On another note, the rates of missing data also raise an important point, namely that even in centres with a focus on the CVD aspect of RA disease, these crucial clinical variables are seldom recorded, thus underlining the importance of increasing the awareness of this clinical field. It should be stressed that one of the objectives of an audit, as opposed to an epidemiological study, is to define the degree of missing information, so that improving data collection becomes part of process improvement. The seemingly poor appreciation of CVD risk in RA patients may be due to low knowledge among health personnel,37 and due to diffusion of responsibility for CVD risk evaluation between general practitioners, cardiologists and rheumatologists. A clinical audit as ours has limitations in the lack of standardized instruments for all centres such as blood pressure monitors, height and weight measuring scales and no central laboratory measurements. Despite this, clinical audits are recognised as a valuable tool in improving clinical performance in implementing guideline recommended procedures.
To be able to compare data from different regions we decided to use the European definitions of CVD risk calculation, although, SCORE is not validated for use outside Europe. This is a methodological limitation. The strengths of this survey are the large number of patients from various countries, and that the collection is from the last decade, which gives up-to-date information.
CONCLUSION
Using data collected in the SURF-RA, we have shown that there is still a high prevalence of CVD risk factors and ASCVD in patients with RA across 3 continents, and that there is an unmet need for vigilance and improved implementation of preventive measures in this high CVD risk patient population. Substantial geographical differences were revealed with regard to prevalence of CVD risk factors and established CVD, and the high risk of CVD in RA patients from Central and Eastern Europe is of special concern. We hope that this effort to present the CVD risk factor burden will serve to inform and support a global community of cardiovascular health personnel in the ongoing quest for improved cardiovascular health in patients with RA. Clinical audits are only of value if process improvement and better outcomes result. Therefore, we suggest: (a) the development of Standard Operating Procedures for risk evaluation and management in RA subjects and (b) periodic re-audits to monitor change in risk factors and prevalence of established CVD.
FUNDING
The work was supported by grants from the South Eastern Regional Health Authorities of Norway (2013064 for A.G.S. and 2016063 for S.R.) and FOREUM (the Foundation for Research in Rheumatology for AMK). Further support was through a collaborative agreement for independent research from Eli Lilly who had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript.
ACKNOWLEDGEMENTS
We are grateful and in depth to all those clinicians and health personnel who have participated in the SURF-RA: Solbritt Rantapää-Dahlqvist,1 George Athanasios Karpouzas,2 Miguel A Gonzalez-Gay,3 Petros P Sfikakis,4 Maria G Tektonidou,4 Argyro Lazarini,5 Dimitrios Vassilopoulos,5 Bindee Kuriya,6 Carol Hitchon,7 Maria Simona Stoenoiu,8 Patrick Durez,8 Virginia Pascual-Ramos,9 Dionicio Angel Galarza-Delgado,10 Pompilio Faggiano,11 Durga Prasanna Misra,12 Andrew A Borg,20 Rong Mu,21 Erkin M Mirrakhimov,22 Diane Gheta,23 Karen Douglas,24 Vikas Agarwal,12 Svetlana Myasoedova,25 Lev Krougly,26 Tatiana Valentinovna Popkova,27 Alena Tuchyňová,28 Michal Tomcik,29 Michal Vrablik,30 Jiri Lastuvka,31,32 Pavel Horak,33 Helena Kaspar Medkova34
1Department of Public Health and Clinical Medicine/Rheumatology, Umeå Universitet Medicinska Fakulteten, Umea, Sweden
2Internal Medicine-Rheumatology, Harbor-UCLA Medical Center, Torrance, California, USA
3Rheumatology, Hospital Universitario Marques de Valdecilla, Santander, Spain
4First Department of Propaedeutic and Internal Medicine, University of Athens, Athens, Greece
52nd Department of Medicine and Laboratory, Clinical Immunology-Rheumatology Unit, National and Kapodistrian University of Athens School of Medicine, Athens, Greece
6Medicine, Mount Sinai Hospital, Toronto, Ontario, Canada
7Rheumatology, University of Manitoba, Winnipeg, Manitoba, Canada
8Rheumatology, Université Catholique de Louvain, Louvain-la-Neuve, Belgium
9Immunology and Rheumatology, Instituto Nacional de Ciencias Medicas y Nutricion Salvador Zubiran, Mexico City, Mexico
10Rheumatology, Hospital Universitario Dr Jose Eleuterio Gonzalez, Monterrey, Nuevo León, Mexico
11Cardiothoracic, Fondazione Poliambulanza Instituto Ospedaliero, Brescia, Italy
12Department of Clinical Immunology and Rheumatology, Sanjay Gandhi Post Graduate Institute of Medical Sciences, Lucknow, Uttar Pradesh, India
13Department of Rheumatology, Mater Dei Hospital, Msida, Malta
14Rheumatology, Peking University People’s Hospital, Beijing, China
15Kyrgyz State Medical Academy Faculty of General Medicine, Bishkek, Kyrgyzstan
16Medicine, Tallaght University Hospital, Dublin, Ireland
17Department of Rheumatology, Dudley Group of Hospitals NHS Trust, Dudley, UK
18Rheumatology, Ivanovo State Medical Academy, Ivanovo, Ivanovskaa oblast', Russian Federation
19Rheumatology, FSBI National Medical and Surgical Center named after N I Pirogov of the Ministry of Healthcare of the Russian Federation, Moskva, Moskva, Russian Federation
20V.A. Nasonova Research Institute of Rheumatology, Moscow, Russian Federation
21Rheumatology, Narodny Ustav Reumatickych Chorob, Piestany, Slovakia
223rd Department of Internal Medicine, General University Hospital and 1st Faculty of Medicine, Charles University, Prague, Czech Republic
23Third Department of Internal Medicine, Department of Endocrinology and Metabolism, Charles University First Faculty of Medicine, Praha, Czech Republic
24Krajska zdravotni a.s, Masaryk Hospital in Usti nad Labem, Usti nad Labem, Czech Republic
25First Medical Faculty, Charles University, Praha, Czech Republic
26Department of Internal Medicine III–Nephrology, Rheumatology and Endocrinology, University Hospital Olomouc, Olomouc, Olomoucký, Czech Republic
27Division of Rheumatology, 2nd Department of Internal Medicine–Gastroenterology, Charles University First Faculty of Medicine, Hradec Králové, Czech Republic
CONFLICT OF INTEREST
The authors report no conflict of interest.
AUTHOR CONTRIBUTIONS
A.G.S. contributed to conception, design, acquisition, analysis, interpretation, and drafted including critically revised and gave final approval of the manuscript; S.R. contributed to conception, design, analysis, interpretation, drafted, critically revised and gave final approval of the manuscript; J.S. contributed to design, acquisition, analysis, interpretation, critically revised and gave final approval of the manuscript; E.I., G.W. and A.M.K. contributed to conception, acquisition, interpretation, critically revised and gave final approval of the manuscript; I.G., P. van R., G.K., C.S.C. contributed to conception, design, interpretation, critically revised, and gave final approval of the manuscript.
DATA STATEMENT
No data are available. All data relevant to the study are included in the article or uploaded as supplementary information. The data were pseudonymised by each centre before transferred to the data handling centre at Diakonhjemmet hospital, where it is stored on a central server. The data handling manager is Joe Sexton (joesexton0@gmail.com) and the data centre leader is Anne Grete Semb (a-semb@diakonsyk.no).