Background Truncating variants in titin (TTNtv) are the most common cause of dilated cardiomyopathy (DCM). We evaluated the genotype-phenotype correlation in TTNtv-DCM, with a special focus on long-term outcomes, arrhythmias, response to treatment and sex-related presentation.
Methods Data on patient characteristics and outcomes were collected retrospectively from electronic health records of patients genotyped at two Danish heart transplantation centres.
Results We included 115 patients (66% men). At diagnosis of DCM, mean age was 46±13 years and left ventricular ejection fraction (LVEF) was 28%±13%. During a median follow-up of 7.9 years, 26% reached a composite outcome of left ventricular assist device implantation, heart transplantation or death. In 20% an arrhythmia preceded the DCM diagnosis. In total, 43% had atrial fibrillation (AF) and 23% had ventricular arrhythmias. Long-term left ventricular reverse remodelling (LVRR; LVEF increase ≥10% points or normalisation) was achieved in 58% and occurred more frequently in women (72% vs 51%, p=0.042).
In multivariable proportional hazards analyses, occurrence of LVRR was a strong independent negative predictor of the composite outcome (HR: 0.05 (95% CI 0.02 to 0.14); p<0.001). Female sex independently predicted lower rates of ventricular arrhythmias (HR: 0.33 (95% CI 0.11 to 0.99); p=0.05), while the location of the TTNtv was not associated with cardiovascular outcomes.
Conclusion DCM caused by TTNtv presented in midlife and was associated with a high burden of AF and ventricular arrhythmias, which often preceded DCM diagnosis. Furthermore, LVRR occurred in a high proportion of patients and was a strong negative predictor of the composite outcome. Female sex was positively associated with occurrence of LVRR and longer event-free survival.
- heart failure
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Statistics from Altmetric.com
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.
Dilated cardiomyopathy (DCM) is a disorder with a wide spectrum of heterogenous aetiologies, defined by systolic dysfunction and dilation of the left ventricle in the absence of abnormal loading conditions or ischaemic heart disease.1 The prevalence of DCM is uncertain but at least 1:2500.2 DCM is an important cause of heart failure, arrhythmias, heart transplantation (HTx) and premature death.3
DCM is classified as familial or non-familial based on occurrence in more than one family member.1 4 The genetic basis of DCM is identifiable in approximately 40% of familial cases and multiple variants in a broad range of genes have been implicated in the pathogenesis.5 6 The mode of inheritance is predominantly autosomal dominant, but autosomal recessive,7 X linked8 and mitochondrial inheritance patterns2 have been described. Truncations of the giant sarcomere protein titin, encoded by the TTN gene, is the most common genetic cause of DCM, with truncating titin variants (TTNtv) found in 15%–25% of DCM cases.9–11
Reports on the DCM phenotype associated with TTNtv (DCM-TTNtv) have shown conflicting results. Some have shown similar clinical manifestations and outcomes in patients with TTNtv-DCM and non-TTNtv-DCM and a worse prognosis for men.11 12 Others have reported a younger age of DCM onset, lower indexed left ventricular mass and stroke volume, but similar biventricular ejection fractions and outcomes when compared with patients with non-TTNtv-DCM.13 Roberts et al found more severe left ventricular impairment, more sustained ventricular tachycardia and worse outcome in TTNtv-DCM.9 Contrary, Dutch and Finnish reports have described favourable outcomes of TTNtv-DCM with less severe presentation and a favourable effect of treatment,14 when compared with patients with LMNA-DCM.14 15
The aim of the current study was to characterise the TTNtv-DCM phenotype, in respect to heart failure, associated arrhythmias, the occurrence of left ventricular reverse remodelling (LVRR), and sex-specific phenotypic differences in a large cohort of TTNtv patients with detailed clinical data and long-term follow-up. Furthermore, we sought to identify the disease-modifying nature of a range of clinical factors on the DCM phenotype.
Study design and population
The study was a retrospective, observational, longitudinal cohort study of patients recruited from two centres in Denmark: The Department of Cardiology, The Heart Centre, Copenhagen University Hospital and The Department of Cardiology, Aarhus University Hospital. These centres provide diagnostic workup, outpatient care and family screening of patients with DCM besides constituting the only heart transplant centres in Denmark. We recruited patients with DCM followed at our clinics from 2006 to 2019. To be included, patients had to carry a heterozygous truncating titin variant (frameshift, splice site or nonsense) and have an otherwise unexplained left ventricular ejection fraction (LVEF) ≤50% as determined by echocardiography.1 Patients with sudden cardiac death, who were found to carry TTNtv and had autopsy features consistent with DCM and living family members with TTNtv-DCM were included.
Patients in which the aetiology to development of DCM was likely to be other causes, such as history of significant coronary artery or valvular disease, severe systemic hypertension, dysregulated diabetes mellitus, side effects to radiation or chemotherapy were excluded. Patients with possible metabolic, infectious or inflammatory cardiomyopathy were also excluded. Both probands and affected relatives were eligible for inclusion. For further information on definitions, please see online supplemental methods.
DNA was extracted from peripheral blood using standard methods. Next-generation sequencing techniques were used for the genetic screening. In Copenhagen, a Haloplex Target Enrichment System (Agilent Technologies, California, USA) was used. DNA was fragmented, hybridised to biotinylated gene-specific probes, captured by magnetic beads, PCR amplified and sequenced on a MiSeq System (Illumina, California, USA). In Aarhus, a KAPA HTP library preparation kit (Qiagen, Hilden, Germany) was used, followed by hybridisation to Nimblegen EZ in solution probes (Roche, Basel, Switzerland), and sequenced on either MiSeq or NextSeq 500 (both Illumina) with 2×150 basepair paired-end reads.
Poor quality reads were filtered, alignment to GRCH37/hg19 was performed with Burrows-Wheeler Alignment tool16 and GATK V.3.4–3.617 was used for realignment, recalibration, duplicate removal and variant calling. A minimum average coverage of 30× was guaranteed (complex genomic regions were Sanger sequenced).
Patients with variants previously reported as pathogenic/likely pathogenic were included. Novel variants were classified according to the American College of Medical Genetics and Genomics (ACMG) guidelines18 and patients carrying these were included if the variant was predicted to result in a loss of function, was rare (GnomAD minor allele frequency <1×10−4) and co-segregated with disease. To further ensure that the studied variants had a high probability of leading to loss of function, only variants in exons previously reported to have a proportion spliced in (PSI) score above 0.9 in cardiac tissue was included.9 The PSI score measures the exon-inclusion ratio, that is, the relative abundance of titin isoforms which contain the exon in cardiac tissue. This cut-off ensures that the likelihood of the affected allele producing a truncated titin-peptide is >90% in the cardiac sarcomeres.
The number of genes screened evolved during the study period. At the end of the study period, the Copenhagen gene panel consisted of 62 genes and the Aarhus gene panel consisted of 102 genes. The initial gene panels in use at our institutions included 14 genes (see online supplemental material for further information). If other genetic variants, classified as pathogenic or likely pathogenic according to the ACMG 2015 guidelines,18 were identified in addition to a TTNtv, these were also registered.
For each included individual we registered age at diagnosis, sex, height, weight, presence of diabetes or hypertension, family history of heart failure, if they were probands, worst New York Heart Association (NYHA) functional class throughout the disease course, LVEF and left ventricular end-diastolic diameter (LVEDD) at the time of diagnosis and whether implantable cardioverter-defibrillator (ICD) or cardiac resynchronisation therapy (CRT) devices had been implanted. We also registered current pharmacological therapy including use of ACE inhibitors/angiotensin-II receptor blockers/combined angiotensin-II receptor blockers and neprilysin inhibitors, beta-blockers, thiazide or loop diuretics and aldosterone antagonists. If a patient had died or had a HTx, we registered the treatment given at the time leading up to death/HTx. We measured PR-interval and QRS-interval from electrocardiograms performed at the latest cardiac assessment (in case of atrial fibrillation (AF) or pace rhythm the latest sinus rhythm ECG was analysed). The presence of AV-block and right/left bundle branch block (RBBB/LBBB) was evaluated at diagnosis of DCM and from the latest ECG. In the subgroup of patients who had clinically indicated myocardial biopsies performed, we registered whether significant interstitial fibrosis had been observed and in patients who had cardiac MRI (cMRI) scans available for analysis, we registered if replacement fibrosis was observed.
Based on the Danish civil registration number, we collected outcome data on all included individuals from electronic health records. The occurrence and date of implantation of left ventricular assist devices (LVAD), HTx and death were registered. Arrhythmic outcomes included development of AF/atrial flutter, other supraventricular arrhythmia, sustained ventricular tachycardia, ventricular fibrillation, aborted sudden cardiac death or appropriate ICD shock. Occurrence of non-sustained ventricular tachycardia or antitachycardia pacing were not registered. Data on arrhythmias were collected from ECG tracings, Holter monitorings, in-hospital telemetry and implantable devices. All performed echocardiograms throughout the disease course were evaluated and the most favourable (highest LVEF and/or lowest LVEDD) and latest echocardiography were registered. Left ventricular reverse remodelling (LVRR) was defined as an absolute increase in LVEF of ≥10% or normalisation (LVEF ≥55%) of systolic function following initiation of heart failure therapy. Arrhythmic events or investigations performed after LVAD implantation or HTx were censored.
Dichotomous variables are presented as frequencies, continuous normally distributed variables as mean±SD and non-normally distributed data as medians with IQR. HRs from regression analysis are presented with 95% CIs. Comparisons between groups were performed using Fisher’s exact test for categorical data, and with Student’s t-test, Welch’s t-test or Mann-Whitney U test where appropriate. The Kaplan-Meier method was used to estimate cumulative freedom from outcomes and the log-rank test was used to test statistical significance. Competing risk analyses was performed, in outcomes in which death, LVAD implantation or HTx were competing events and statistical significance was tested with a likelihood ratio test.
Predictive clinical models for the combined outcome (LVAD implantation, HTx or death), malignant ventricular arrhythmias and AF as a function of follow-up was created. Univariate Cox proportional hazards regression analyses were performed to assess covariates associated with event-free survival. Multivariable Cox proportional hazards regression analyses were performed using variables associated with event-free survival corrected for random effects caused by relatedness. Variables known to be clinically relevant were added to the model and further variables were added to the model in a stepwise (forward) manner with p<0.1 for inclusion and p<0.05 for retention (see online supplemental methods). All tests were two-sided, and the level of significance was set at a p value <0.05. Statistical analyses were conducted in Excel (Microsoft, Washington, USA) and in R, V.3.6.1 (R Foundation for statistical computing, Vienna, Austria).
In total, we identified 115 patients from 80 families (66% men; 99% Northern European decent) with predicted TTNtv-DCM eligible for inclusion. See figure 1 for information on reviewed patients and table 1 for patient characteristics.
We identified 54 heterozygous frameshift (n=27), nonsense (n=22) or splice (n=5) variants predicted to lead to truncations of titin (online supplemental table 1). All nonsense and frameshift variants were in exons with a PSI score of 1.0, meaning they were constitutively expressed in cardiac muscle. Three of the splice variants were predicted to alter splicing of constitutively expressed exons while the last two variants co-occurred (compound heterozygotes) and segregated with DCM in a single family. Thirty-five of the described variants have not been reported previously. Twenty-four variants were found in more than one of the included subjects totalling 85 patients (74%). In contrast, 30 variants were private variants and a subset of these could be de novo variants, although none was positively identified as such, and 18 of these patients reported occurrence of heart failure in first-degree relatives. Variants were more commonly located in the more distal segments of titin, that is, at a distance from the large areas of the I-band which undergoes alternative splicing (figure 2); most frequently in the A-band (53%), followed by the I-band (23%), and M-band (20%), while truncating variants in the Z-disk were rare (3%). Secondary genetic variants identified as likely pathogenic were found in SCN5A (NM_198056.2:c.638G>A, p.Gly213Asp) in two related patients and in DSG2 (NM_001943.5:c.137G>A, p.Arg46Gln) in one patient (see online supplemental material S1 for clinical description of these patients).
Clinical characteristics associated with TTNtv
Mean age at diagnosis of DCM was 46±14 years; 62% were probands, and a family history of heart failure/DCM was frequent (82%). Diabetes and hypertension were diagnosed in 8% and 13%, respectively. At diagnosis mean LVEF was 28%±13%, and LVEDD 32±5 mm/m2. Almost half (48%) of patients had at some point registered a NYHA functional class of III–IV. Most patients tolerated standard heart failure medications (see table 1). ICDs were implanted in 22 (19%) patients while CRT was infrequently used (8% for CRT-D; 2% for CRT-P). In the subgroup of patients who had myocardial biopsies performed (n=27), significant interstitial fibrosis was found in 20 (74%) patients and in the subgroup of patients who had cMRI scans performed (n=12), three (25%) patients had evidence of replacement fibrosis as evidenced by late gadolinium enhancement (n=2) or hinge-point fibrosis (n=1). Analyses of the latest ECG showed a mean PR-interval of 182±36 ms and a mean QRS duration of 105±22 ms. At this point, first-degree AV-block occurred in 30%, QRS duration >120 ms in 20% and QRS duration >150 ms in 8% of patients. If evaluating results from both the first and latest ECG the prevalence of AV-block was 36%, while 22% had a LBBB or RBBB. Results are summarised and subgrouped by sex in table 1.
Men were significantly younger at diagnosis of DCM, with a median age of 44 years vs 52 years in women (figure 3A). No significant sex difference was observed in LVEF or indexed LVEDD at diagnosis (both p>0.05). At the latest performed ECG, no difference was observed in incidence of AV-block or LBBB or RBBB (both p>0.05).
Heart failure outcomes
During a median follow-up of 7.9 years (IQR: 3.3–13.3 years), 30 patients (26%) reached a combined outcome of HTx (n=18), LVAD implantation (n=5) or death (n=20).
Men reached the combined outcome more frequently (table 2) and at a significantly younger age than women (figure 3B). In multivariable mixed modelling the occurrence of LVRR during follow-up was an independent positive predictor of freedom from the combined outcome (HR: 0.05 (95% CI 0.02 to 0.14); p<0.001). Other factors included in the model did not independently predict the combined outcome; age at DCM diagnosis (HR: 1.07 (95% CI 0.89 to 1.28) per 5 years; p=0.484), LVEF at diagnosis (HR: 0.88 (95% CI 0.71 to 1.09); p=0.225) or being a proband (HR: 2.14 (95% CI 0.65 to 7.07); p=0.210). Results from univariable Cox regression analysis and visualisation of results for the combined outcome can be found in online supplemental table 2 and online supplemental figures 1 and 2.
Survival analyses found patients with TTNtv in the Z-disc or I-band were significantly older than patients with TTNtv in the A-band at diagnosis of DCM (online supplemental figure 3A) while the location of the TTNtv was not associated with any other outcome (online supplemental figure 3B–D and online supplemental tables 2–4). As expected, being a proband was associated with a younger age at occurrence of the combined outcome (online supplemental figure 4B).
During follow-up, the overall occurrence of arrhythmias was high (58%), AF occurred in 50 patients (43%), ventricular arrhythmias in 26 (23%) and other supraventricular arrhythmias in 15 (13%) patients. Men had a higher prevalence of ventricular arrhythmias (table 2) and presented with a ventricular arrhythmia earlier in life than women (figure 3C). Furthermore, men were significantly younger at occurrence of any arrhythmia (median age 52 years in men vs 67 years in women (p=0.003)). Survival analysis in probands versus relatives found the occurrence of ventricular arrhythmias was associated with being a proband (online supplemental figure 4C), while this was not the case in the AF outcome (online supplemental figure 4D).
An arrhythmia preceded the DCM diagnosis in 23 (20%) patients and was diagnosed concurrently with (n=6) or within 2 weeks (n=7) of diagnosis of DCM in 11% of patients. In 18 (16%) patients, AF preceded the diagnosis of DCM and in 11 (10%) patients, AF was diagnosed concurrently (n=9) or within 2 weeks (n=2) of DCM. In 13 (11%) patients, the presenting symptom of DCM was a ventricular arrhythmia of which two (2%) had a fatal outcome.
Multivariable Cox regression analysis of the ventricular arrhythmia outcome found male sex (HR: 3.05 (95% CI 1.00 to 9.32); p=0.050), being a proband (HR: 5.62 (95% CI 1.61 to 19.61); p=0.007) and LBBB/RBBB at baseline (HR: 2.44 (95% CI 1.05 to 5.71); p=0.039) to be independent positive predictors of ventricular arrhythmia, while the age at diagnosis of DCM (HR: 1.03 (95% CI 0.86 to 1.25) per 5 years; p=0.740) was not associated with this outcome (online supplemental table 3).
In multivariable testing of the AF outcome, AV-block at baseline (HR: 2.51 (95% CI 1.22 to 5.16); p=0.012) was an independent positive predictor of the outcome while the two predefined clinical variables: age at DCM diagnosis (HR: 1.12 (95% CI 0.99 to 1.27); p=0.074) and being a proband (HR: 0.97 (95% CI 0.48 to 1.96); p=0.930) did not independently predict rates of AF (online supplemental table 4).
Survival analyses did not find the occurrence of AF to be associated with the occurrence of the LVAD, HTx or death outcome.
Reversibility of systolic function
Data from at least two echocardiographic examinations following diagnosis of DCM were available in 108 patients. At the latest echocardiography, performed after a median of 6.0 years (IQR: 1.6–10.7 years) of follow-up, 58% of patients met criteria for LVRR. At this timepoint, LVEF had improved from 28±13% to 39±16% (p<0.0001) and LVEDD had decreased from 32±5 to 30±4 mm/m2 (p<0.0001). Results are summarised in table 2 and figure 4.
Left ventricular reverse remodelling occurred more frequently (72% vs 51%, p=0.042) and was achieved faster (median 6.7 vs 19.3 months, log-rank p=0.0002) in women; at the latest performed echocardiography LVEF was higher in women (43% vs 37%, p=0.039).
This study investigated the phenotype of TTNtv-DCM in a large Danish cohort with long-term follow-up. We reported TTNtv in all regions of titin, however with a distribution skewed towards the C-terminal as reported previously.9 11 13 15 19–21 Most patients had an onset of DCM between the fourth and sixth decade of life. The disease course was characterised by a high burden of both atrial and ventricular arrhythmias. Remarkably, an arrhythmia was the presenting symptom in 31% of patients. While 26% of patients reached a combined outcome of LVAD, HTx or death during follow-up, a high proportion of patients had LVRR at the end of follow-up and occurrence of LVRR was a strong independent positive predictor of event-free survival, reducing the risk of reaching the combined end point more than 10-fold. Finally, a marked difference in onset of DCM, disease progression, and response to treatment was observed between sexes. Female sex was associated with lower rates of ventricular arrhythmias, approximately by a factor of 3, independently of other disease-modifying factors.
End-stage heart failure
The prevalence of LVAD, HTx or death was approximately twice as high in our cohort when compared with previous studies.9 13 14 However, this finding is not remarkable in DCM22 and is likely explained by the long follow-up time in our study and recruitment bias. Nonetheless, the proportion of patients requiring advanced heart failure treatment does indicate that some patients reach a ‘point-of-no-return’ in which disease progression can no longer be reversed or delayed by standard heart failure treatment alone. This finding highlights the importance of early recognition of DCM and initiation of therapy.
The prevalence of AF was 43% in our cohort which is up to two times higher than previously reported in TTNtv-DCM13 14 and two to four times higher than in general DCM.3 23 24 Interestingly, 23% of patients developed AF before or concurrently with a DCM diagnosis. These findings are consistent with case-control studies describing TTNtv as an important risk factor for early onset AF.25 26 Importantly, our data do not support presence of AF as a marker of poor prognosis and LVRR was similar between patients with or without AF.
Ventricular arrhythmias were common in our cohort (prevalence of 23%) and was the first disease manifestation in half of these cases. Ventricular arrhythmias have previously been observed to be more common in TTNtv genotype/phenotype studies compared with non-TTNtv-DCM,9 13 24 27 although less common than seen in LMNA-DCM.14 28 The prevalence of ventricular arrhythmia in TTNtv-DCM has not previously been reported in larger phenotype studies, but the prevalence we report is similar to previous reports in a smaller patient cohort.29
We observed significantly lower event rates of ventricular arrhythmia in women compared with men. This finding was independent of other possible disease-modifying factors. If replicated, this finding could have important clinical implications when selecting patients for prophylactic ICDs.
While studies support TTNtv to be an independent predictor of early and life-threatening arrhythmias in DCM,27 29 the molecular mechanisms behind the development of arrhythmias in TTNtv-DCM are incompletely understood. Rodent and zebrafish models25 30 31 document development of diastolic dysfunction and cardiac fibrosis in both atria and ventricles which may serve as a substrate for both atrial and ventricular arrhythmias.32 33 Furthermore, endomyocardial biopsies from patients with TTNtv-DCM reveal increased interstitial fibrosis compared with patients with DCM from other causes.29 A more diffuse pattern of myocardial fibrosis could potentially explain the high burden of arrhythmias and is supported by findings from myocardial biopsies in our study.
Left ventricular reverse remodelling
In a large proportion of the patients, significant improvements in cardiac function meeting LVRR criteria was observed and improvements were largely sustained long-term. This finding is surprising considering the severe heart failure phenotype observed at baseline. Patients with milder phenotypes at DCM diagnosis could have an even better recovery of systolic function and better cardiac outcomes, as previously indicated when comparing response rates with pharmacological therapy in patients with non-ischaemic heart failure.34
Previous smaller studies have investigated recovery of cardiac function in TTNtv in other clinical settings and found indications that the TTNtv-DCM disease-course is milder when compared with other genetic causes of DCM.14 28 34 Felkin et al found a similar recovery in patients with severe non-ischaemic DCM treated with an LVAD with or without TTNtv.35 A Japanese and a Dutch study investigated the prognosis and LVRR in patients with DCM and found TTNtv-DCM compared favourably with other DCM groups in terms of rate of LVRR.28 34 Importantly, occurrence of LVRR was in our study a strong predictor of lower rates of LVAD, HTx or death, independently of other disease-modifying factors.
Prior studies have observed male patients with TTNtv to have a more severe phenotype, with an earlier onset of DCM and worse outcomes.11 12 Further supporting a more severe phenotype in men with TTNtv, is the 2:1 male-to-female ratio in our cohort, a finding consistent with prior studies of TTNtv-DCM11–14 28 29 and in DCM in general.36 Considering an autosomal dominant inheritance pattern in TTNtv, a skewed sex ratio of disease is surprising and suggests secondary genetic, post-transcriptional, environmental, hormonal and other factors modifies disease progression. Indeed, prior studies support this concept of disease modifiers in TTNtv carriers.37–39
Interestingly, women who had manifest DCM exhibited better recovery of systolic function in our cohort than men, a finding not previously reported. While this is in line with a less severe phenotype, a previous study investigating clinical predictors of LVRR in patients with DCM found female sex to be associated with lower ratios of LVRR.34 A higher proportion of LVRR in women versus men might relate to hormonal factors, reduced compliance or high-risk behaviour in men, by more advanced disease at time of diagnosis or a higher baseline level of haemodynamic stress in men (higher cardiac output, cavity dimensions and blood pressure). Importantly, a larger proportion of men were probands, and although the difference was not statistically significant, it could impact some of the studied outcomes.
Significant sex differences in phenotypes have been reported in patients with DCM, notably when caused by variants in LMNA and RBM20.40 41 In both LMNA and RMB20, men are younger at disease onset and have worse outcomes. The sex differences observed in RBM20 cardiomyopathy is especially relevant in the context of this study, since RBM20 regulates titin splicing.42
Our study cohort were included from the two national HTx centres introducing a selection and referral bias towards patients with more severe phenotypes, earlier onset of disease, less comorbidity and prominent family histories. By study design, data were collected retrospectively with its caveats. While one of the strengths of our study was a long follow-up time, it introduces a bias in terms of treatments offered. As an example, the rate of implanted ICDs is relatively low in our cohort since we have been restrictive in implanting ICDs in patients with non-ischaemic cardiomyopathy in Denmark.3 Furthermore, few patients received a neprilysin inhibitor or ivabradine. We cannot rule out that these newer therapies could have changed outcomes.
TTNtv led to a DCM phenotype characterised by a high burden of both AF and ventricular arrhythmias, which often precedes a DCM diagnosis. While 26% reached a composite outcome of LVAD, HTx or death during 7.9 years of follow-up, a significant proportion of the patients improved markedly in response to heart failure treatment and occurrence of LVRR was a strong positive predictor of a better long-term outcome. Finally, female sex was found to independently predict lower rates of malignant ventricular arrhythmias and women had a better response to medical treatment, were older at onset of DCM and at end-stage heart failure outcomes.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Patient consent for publication
The Danish Data Protection Agency and Patient Safety Authority approved the study, and authors had full access to the data. The study followed the Helsinki Declaration.
Contributors Concept and design: AHC, HB and CRV. Acquisition of data: CRV, TBR, AMD, MSO, LNP, MJ and AHC. Statistical analyses and interpretation of data: CRV, HB and AHC. Drafting of manuscript and critical revision of the article: CRV, TBR, AMD, MSO, LNP, MJ, HB and AHC.
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.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.