ABSTRACT
The diagnostic yield of endomyocardial biopsy in cardiac sarcoidosis (CS) is quite low because of the patchy involvement, and for the diagnosis of CS, existing guidelines required histological confirmation. Therefore, especially for isolated CS, diagnosis consistent with the guidelines cannot be made in a large number of patients. With recent developments in imaging modalities such as cardiac magnetic resonance and 18-fluorodeoxyglucose positron emission tomography, diagnosing CS has become easier and diagnostic criteria for CS not compulsorily requiring histological confirmation have been suggested. Despite significant advances in diagnostic tools, large-scale studies that can guide treatment plans are still lacking, and treatment has relied on the experience accumulated over the past years and the consensus of experts. However, opinions vary, depending on the situation, which is quite puzzling for the physician treating CS. Moreover, withM
the advent of new immunosuppressant agents, these new drugs have been applied under the assumption that the effect of immunosuppression is not much different from that of other well-known autoimmune diseases that require immunosuppression. However, we should wait to see the beneficial effects of these new immunosuppressants before we attempt to apply these agents in our clinical practice. This review summarises the widely used diagnostic criteria, current diagnostic modalities and recommended treatments for sarcoidosis. We have added our opinions on selecting or modifying diagnostic and treatment plans from the diverse current recommendations.

INTRODUCTION
Sarcoidosis is a granulomatous disease of unknown aetiology characterised by multisystem involve- ment. Among the organs involved in systemic sarcoidosis, the lung is by far the most frequently involved (>90%), and cardiac involvement is noted only in 2%–3% of the patients.^{1 2} While this low incidence of cardiac sarcoidosis (CS) reflects the low incidence of cardiac involvement in systemic sarcoidosis, a lack of eagerness to diagnose CS may also play a role.
  Previously, myocardial thickness and motion assessed by echocardiography were the only param- eters available for the detection of myocardial injury associated with inflammation. With the evolution of imaging modalities, such as cardiac magnetic reso- nance (CMR) and 18-fluorodeoxyglucose–positron emission tomography (FDG-PET), information about myocardial injury or inflammation may be obtained in the early phase of CS.
  From a therapeutic perspective, many new immu- nosuppressants have been introduced. Moreover, the roles of current interventional therapies, such as ablation therapy for tachyarrhythmia and implant- able cardioverter defibrillator (ICD) implantation, remain to be defined.

EPIDEMIOLOGY
Although infrequently affected, the heart is an important site of involvement, and a major deter- minant of prognosis. The first case of heart involvement in sarcoidosis was reported by Bern- stein in 1929, and clinically apparent CS has been noted in approximately 5%–10% of patients with systemic sarcoidosis.^{1 3} However, a much higher rate of myocardial involvement of approximately 20%–30%⁴ to over 70%⁵ has been reported in autopsy series. With the development of imaging modalities and clinical awareness of CS, the inci- dence of the diagnosis of cardiac involvement in systemic sarcoidosis may reach that of autopsy series.

PATHOGENESIS
Sarcoidosis is a multisystem granulomatous disease characterised by the presence of non-caseating granulomas in the involved organ.^{6 7} In granuloma- tous disease, granuloma formation is thought to be a host defensive mechanism to prevent the spread of pathogens or irritant events.⁸    Pathological processes are believed to start with circulating monocytes that interact with a specific antigen and differentiate into specific antigen- presenting cells (APCs). In this process, in the pres- ence of cytokines, APC promotes CD4+ T cells expressing T-helper (Th)-1, Th-17 and Th-17.1 cells. By secreting their specific cytokines, corre- sponding responses are elicited and inflammation in the affected sites is promoted (figure 1). Gener- ally, a regulatory T (Treg)-cell response is known to repress local helper T-cell responses and has anti- inflammatory action. In sarcoidosis, although not always pronounced, extensive local inflammation is associated with peripheral anergy, which is called the ‘immune paradox’. Several studies have shown that Treg cells are amplified at the lesion site and as well as in the peripheral blood but are still unable to control local inflammation as tumour necrosis factor (TNF)- is weakly inhibited but plays a role in granuloma formation at the lesion site.⁹

This simple description of immunopathogenesis is to help understand our therapeutic strategy. The important role of Th-17 cells in the pathogenesis, and CD+ T-cell plasticity were only recently appre- ciated. Much is still unknown about the role of Treg and serum amyloid A protein in sarcoidosis.

   Considering the role of APC in initiating this process, it is natural to consider the role of major histocompatibility complex (MHC) class II mole- cules, normally found only on APC. In humans, the MHC class II protein complex is encoded by the human leucocyte antigen (HLA) gene complex, and an association between genetic susceptibility and HLADQ− and DR− genes has been reported. Associations between non-HLA genes, such as the butyrophilin like 2 (BTNL2) gene, located close to the MHC II region, and predisposition or susceptibility to sarcoidosis have also been reported.
   In the therapeutic concept, many different kinds of cytokines are involved in the pathological process, and second-line or third- line drugs have different targets¹⁰ in their anti-inflammatory actions (figure 1).

Figure 1Immunopathogenesis and treatment targets. On antigen presentation to a TCR on a T lymphocyte via MHC class II molecules, various cytokines, chemokines and other soluble mediators are produced, and polarisation of T cell into Th1, Th17 and Th17.1 cells is promoted. Th1, Th17 and Th17.1 responses in affected sites together with the impaired Treg cell response allows enhanced local effector T-cell responses to persist, resulting in chronic sarcoidosis. APC, antigen-presenting cell; MHC, major histocompatibility complex; TCR, T-cell receptor; Th, T helper; Treg, regulatory T.

DIAGNOSIS
Diagnostic criteria
Two major diagnostic criteria for CS have been widely cited and applied in clinical practice. The first is the criteria originally suggested by the Japanese Ministry of Health and Welfare11 in 1993 and later revised by Japanese organisations in 2006, 2014 and 2015^12–14 (online supplemental table 1). The second is the criteria included in the consensus statement from the Heart Rhythm Society (HRS)¹⁵ in 2014 (online supplemental table 2). According to these criteria, when sarcoidosis is not histologically confirmed in the heart, the presence of extra-CS should be docu- mented for the diagnosis of CS.
   Due to the patchy nature of involvement in CS, the diagnostic yield of endomyocardial biopsy (EMB) has been reported to be approximately 20%.^{16 17} Even with electrogram-guided EMB,¹⁸ non-caseating granuloma was noted in 5/13 specimens from three patients who were finally diagnosed as CS.    Because of the difficulty in the histological confirmation of non-caseating granulomas in the heart, even with compelling clin- ical evidence of CS, the diagnosis of extra-CS has become a top priority. However, there are cases of isolated CS^{19 20} in which the diagnosis of extra-CS cannot be made. Although it is debatable whether isolated CS represents isolated involvement of the heart for the entire course of the disease,²¹ the existence of isolated CS, at least at a certain time point in the course of disease, is not in question. In 2016, the Japanese Circulation Society (JCS) updated the guidelines for diagnosing and treating CS.¹⁴ This guideline includes diagnostic criteria for isolated CS (table 1A). In essence, isolated CS can be diagnosed without demonstrating extra-CS and in the absence of histological confirmation in biop- sies from the heart, when one absolute criterion of radionuclide imaging and an additional three other major criteria in their previous criteria are satisfied. However, considering that FDG- PET can show positive results in non-inflammatory conditions as well as in situations when myocardial FDG uptake is inad- equately suppressed, FDG-PET as an absolute criterion might lower the sensitivity or specificity of the criteria. We previously proposed a scoring system for CS (table 1B).²² Compared to the 2016 JCS criteria for isolated CS, our scoring system simply gave more weight to FDG-PET findings instead of regarding FDG- PET findings as an absolute criterion, excluded the left ventric- ular (LV) dysfunction not associated with regional wall motion abnormality (RWMA) unusual for the coronary artery territory and atrioventricular (AV) conduction disturbance, and included the possible accompanying clinical findings.

Electrocardiogram and Holter monitoring
Two clinically important findings focused on by many clinicians are:(1) conduction disturbance and (2) ventricular tachycardia (VT). As the predilection site of CS is the basal anterior septum, the infra-His block is more likely to be expected in cases of complete AV block (figure 2). Therefore, in treating premature ventricular or atrial contraction, which are frequently encoun- tered in patients with sarcoidosis, beta blockers should be care- fully reconsidered in patients with first-degree AV block as it can represent trifascicular block.    In patients with CS, epsilon waves are reported and do not necessarily indicate the diagnosis of arrhythmogenic right ventricular dysplasia (ARVD).²³ However, this phenomenon should be interpreted considering that differentiating between CS and ARVD is difficult. Holter monitoring is frequently performed to detect the presence of VT even in patients without symptoms of palpitation.

Figure 2 (A) Resting EKG shows first-degree AV block with wide QRS complex. As a cause of PR prolongation, the possibility of trifascicular block should be considered. (B) Intermittent complete AV block noticed during Holter monitoring shows wide QRS escape beats suggesting infra-His block. (C) Even if only high echogenicity is noted at the basal anteroseptal segment, certain pathology is likely to be present in the presence of infra-His AV block (online supplemental video) (D) Therefore, the tiny mid-wall delayed enhancement at the basal anterior septum (circle), which can be ignored in other cases, is an important finding in this case. AV, atrioventricular.

Echocardiogram
The typical echocardiographic finding is multiple regional wall motion abnormalities (RWMA) that do not match the coronary artery territory. Multiple patchy sarcoid involvement does not always result in multiple RWMA. Moreover, RWMA unusual for coronary artery territory is difficult to determine, even for an expert echocardiographer. CT coronary angiography is helpful in this situation, that is, RWMA in patients with normal coro- nary arteries.

Among RWMA, thinning of the basal anterior septum seems to be rather specific to CS, although the exact specificity of this finding has not been elucidated (figure 3). However, if only this finding is considered in echocardiographic diagnosis, many patients will be missed, as the reported incidence of this finding is only 20%.
  Cyclic variation of integrated backscatter or longitudinal strain to increase the sensitivity CS detection was suggested. However, considering the limited additive effect of these modal- ities in increasing sensitivity or specificity, it would be better to resort to other imaging modalities such as CMR or FDG-PET in patients with suspected CS showing subtle echocardiographic abnormalities.
  In addition to the role of echocardiography in diagnosing CS, echocardiography is an essential tool in monitoring LV function in patients with CS.
   Several studies^{24 25} have reported diagnostic ambiguities between CS and ARVD. Therefore, in cases where echocardi- ography is suggestive of ARVD, the possibility of CS should be considered, and histological or genetic evidence should be sought for differential diagnosis. In one patient with CS who under- went cardiac transplantation, serial echocardiography showed that the right ventricular (RV) morphology became similar to that of ARVD over time, and histological examination of the explanted heart showed replacement by fatty tissue instead of fibrosis (figure 4).

Figure 3 Involvement of basal anteroseptal segment, which is a relatively unique feature of cardiac sarcoidosis. (A) Fibrosis without wall thinning, (B) endocardial fibrosis with wall tinning, (C) diffuse septal thinning and (D) localised septal thinning.

Figure 4 (A,B) Difference in echocardiographic features 16 months apart. In addition to the slightly enlarged right ventricle, the RV apex showed aneurysmal bulging during systole (arrowheads), mimicking ARVD. (C,D) Explanted heart after cardiac transplantation. The myocardium stained blue (Masson trichrome stain), suggesting focal fibrosis. In addition, large areas of fatty tissue replacement are seen in the right and left ventricles. ARVD, arrhythmogenic right ventricular dysplasia; RV, right ventricular.

CT coronary angiogram
Because of the presence of RWMA, coronary angiography is not infrequently performed in patients with CS. Only to confirm the RWMA which is unusual for the coronary artery territory, CT coronary angiography is sufficient to achieve the goal. In addition, on the CT coronary angiogram, we can confirm the presence of LN’s that are accessible to endobronchial ultrasound (EBUS)-guided biopsy. Therefore, the lower para-tracheal LN should be covered in CT coronary angiography (figure 5A).

Cardiac magnetic resonance
Based on the pattern of delayed enhancement (DE) on CMR, differentiation between ischaemic and non-ischaemic cardio- myopathies can be made quite easily²⁶ (figure 5B). A variety of lesions can be classified as non-ischaemic cardiomyopathy. However, if the non-ischaemic pattern of DE is multiple and patchy in nature, we cannot think of any disease other than CS. Moreover, it is not unusual to observe DE in segments without an RWMA. Therefore, DE evaluation seems more sensitive than the other methods used in the past to detect cardiac involvement in sarcoidosis.
   Applications of various techniques in CMR have been studied in CS, such as T2-weighted images or T1 mapping. However, despite the large number of publications suggesting the useful- ness of these techniques, whether these techniques significantly upgrade the usefulness of CMR in real-world practice remains unclear.
In addition to its diagnostic utility, RV involvement can be assessed in CMR.²⁷ It is reasonable to think that RV involvement can be detected in the early stage by RV DE compared with RV dilatation on echocardiography.
  The prognostic value of CMR imaging in patients with CS has also been studied. In addition to LV DE, RV DE in patients with CS is reported to be associated with a risk of adverse events, particularly ventricular tachyarrhythmias.²⁸

Figure 5 (A) CT coronary angiogram can demonstrate normal coronary arteries in patients with RWMA, therefore suggesting the non-ischaemic nature of RWMA. In addition, the presence of hilar LNs accessible to EBUS biopsy (arrows) can be assessed. Therefore, CT coronary angiogram has dual objective. (B) Cardiac magnetic resonance shows subepicardial DE at the anterior segment (arrowheads) and mid- wall DE at inferoseptal and inferior segments (arrows). Subepicardial or mid-wall DE indicates non-ischaemic cardiomyopathy. LAD, left anterior descending artery; LCx, left circumflex artery; RCA, right coronary artery; DE, delayed enhancement; EBUS, endobronchial ultrasound; RWMA, regional wall motion abnormality.

18-Fluorodeoxyglucose–positron emission tomography
Several studies have shown that FDG-PET is more sensitive than thallium or gallium scans for the diagnosis of CS. Therefore, FDG-PET has almost completely replaced other agents, espe- cially as a fusion image with CT (FDG-PET/CT). In FDG-PET, the normal myocardium’s use of glucose as an energy source should be adequately suppressed. For this purpose, prolonged fasting, a high-fat diet before prolonged fasting, and preadminis- tration of heparin have been used.
   As FDG-PET findings provide not only the location but also the activity of the disease, we therefore assigned more weight to the FDG-PET results in our suggested criteria, and FDG-PET findings are regarded as an absolute criterion in the diagnosis of isolated CS in the JCS updated guidelines. In addition to its role in diagnosis, FDG-PET is used to monitor treatment response (figure 6A).
  To overcome the subjectivity in visual estimation, various methods based on standardised uptake values (SUVs), including SUVmax, SUVmean, cardiac metabolic activity, cardiac meta- bolic volume and coefficient of variance, have been proposed.²⁹

Figure 6 (A). FDG-PET in the assessment of treatment response. FDG-PET image findings before (left column) and after (right column) treatment with steroid. (B) A patient presented with cardiac arrest and in whom ICD was implanted. FDG-PET immediately after ICD implantation (above). Refused taking steroid after a couple of days of treatment complaining of multiple symptoms, and the patient was not regularly followed up after discharge. FDG-PET done 2 years later (below) showed similar uptake in intensity and distribution compared with the previous study. There was no significant change in LV function between two examinations. Hot uptake in FDG-PET image does not necessarily indicate presence of active inflammation. FDG- PET, 18-fluoro-2-deoxyglucose–positron emission tomography; LV, left ventricular.

Serum markers
Elevated serum ACE activity in sarcoidosis has been well known and was frequently used for diagnosis. Although the ACE level in sarcoidosis might be higher than that in normal subjects, the sensitivity of the ACE level for detecting sarcoidosis precludes it being used as a diagnostic test.
Troponin has been proposed for monitoring the disease activity in patients with CS; however, there are still insufficient data to define the clinical usefulness in CS. The role of other biomarkers, including C reactive protein, in sarcoidosis has been studied; however, their role in CS remains unclear.

TREATMENT
Usefulness of steroid or immunosuppressant treatments for CS
In a meta-analysis³⁰ regarding corticosteroid and immunosup- pressant therapy, we can recognise that data regarding its effect on mortality were too limited to conclude that steroid therapy truly reduces mortality rate in CS. Yazaki et al³¹ showed better long-term survival in patients using steroids; however, patients whose CS was diagnosed at autopsy were selected as a control group of non-steroid user. In one recent study, in contrast, the mortality was lower in patients without immunosuppressive therapy (1/21, 4.8%) compared with the group with immuno- suppression (6/70, 8.6%).³²
  The effect on LV function was not consistent in the meta- analysis. In a study by Nagai et al,³³ the composite endpoints of all-cause death, symptomatic arrhythmias and heart failure requiring admission were reduced with steroid therapy. However, no significant differences were found in terms of cardiac death or symptomatic arrhythmias between steroid users and non-users, which seems to indicate that the main effect of steroid therapy is reducing heart failure requiring admission by preventing the deterioration of LV function.
   A relatively consistent beneficial effect of steroid or immuno- suppressant therapy is the recovery of AV conduction. In roughly half of the patients with high-grade AV block, recovery of AV conduction is expected.³⁰
   The major clinical effect of glucocorticoids is the anti- inflammatory activity. However, glucocorticoid therapy is a reasonable option in situations where fibrotic lesions are the dominant finding, expecting an antifibrotic effect. Among diseases treated with steroids, both active inflammation and fibrosis are harmful reactions; therefore, the use of steroids is not contradictory. However, granulomas represent an inflam- mation confined to the centre, with fibroblast and collagen encasing and restricting inflammation, thereby protecting the surrounding tissue. Therefore, the fibrosis-preventing effect of steroids might help prevent progressive fibrosis later in the course of the disease but might lessen the protective effect of granulomas. Therefore, it is desirable to search for active inflam- mation before starting steroid or immunosuppressant therapies. Hot spots on FDG-PET have been regarded as evidence of active inflammation; however, a hot spot in FDG-PET can be seen in conditions other than inflammation. Therefore, in our insti- tution, we start steroid or immunosuppressant therapy when additional clinical evidence of active inflammation, such as new RWMA, recently deteriorated LV function or newly appeared conduction disturbance is also present. In addition, steroid or immunosuppressant therapy might be limited to patients whose LV function is relatively preserved (EF >35%), as the expected effect of these therapies is to preserve LV function, not recover already deteriorated LV function, and they seem ineffective in preventing malignant ventricular arrhythmia in patients with severely depressed LV function.

Initial steroid dose and treatment strategy
In contrast to pulmonary sarcoidosis, high-dose prednisolone (1 mg/kg/day) was recommended as the initial dose³⁴ in cardiac or neurosarcoidosis expecting minimal treatment failure with steroid therapy. However, there has been no general consensus regarding this high initial dose of prednisolone. In one study,³¹ the overall survival of patients treated with a high initial dose (>40 mg/day) did not differ from that of patients treated with a low dose (<30 mg/day). In this study, dose comparison was performed on only 75 patients, and the survival curves between the high-dose and low-dose groups showed differences in the early phases (2–4 years), with similar survival rates after 5 years. Although this study has shortcomings, many patients, especially female patients, cannot tolerate high doses of steroids for a long period. Hence, most experts suggest a low initial dose (0.5 mg/ kg/day of prednisolone) of steroids.⁷   The currently suggested schemes of steroid treatment in CS, in large part, have their basis on the schemes used in the treatment of pulmonary sarcoidosis.³⁵ A low initial dose of steroids is main- tained for 2–3 months, after which the therapeutic response is evaluated with FDG-PET. Once there is a therapeutic response, the dose of the initial steroid therapy is tapered down to the maintenance dose for 3 months, and the final dose is maintained for 9–12 months.⁷ In our institution, we use the same initial dose of steroid for 1 month, tapering to the maintenance dose during the next month, and then the maintenance dose for the next 10 months, with a total duration of treatment of 12 months.
  As with the use of steroids in other autoimmune diseases, the side effects of steroids should be carefully monitored. Consid- ering the risk of steroid-induced avascular necrosis of the femoral head, the patient’s age and occupation should also be considered before starting steroid treatment.

Treatment strategy in patients unresponsive to initial steroid therapy
Patients who are non-responsive to initial steroid therapy must be separated from those who respond to initial steroid therapy but fail to taper the steroid to the maintenance dose level (<10 mg/ day). If an equivalent dose of more than 10 mg/day of prednis- olone is required as the maintenance dose, second-line drugs, such as methotrexate or azathioprine, as a corticosteroid-sparing agent are usually considered. Clinical data regarding initial treat- ment with second-line or third-line drugs in patients who are intolerant of initial steroid therapy are lacking.
  Non-responsiveness to the initial steroid therapy is usually assessed using FDG-PET. Even with quantitative assessment, evaluation of steroid response by FDG-PET is not always straightforward. Therefore, simultaneous clinical assessments, such as improvement in haemodynamic status or conduction disturbances, should also be considered while evaluating unre- sponsiveness (figure 6B).
  The most commonly recommended third-line drug is a TNF inhibitor. TNF is a proinflammatory molecule; therefore, TNF inhibitors have been used to treat various autoimmune and inflammatory diseases. However, preclinical and clinical data have shown that it might show a paradoxical anti-inflammatory and immunomodulatory effects.³⁶ Studies examining TNF--de- ficient animals have reported a critical role of TNF- in regu- lating and limiting the extent and duration of the inflammatory response.³⁷ Therefore, the effectiveness of TNF inhibitors in other autoimmune diseases may not automatically be extended to the effectiveness of sarcoidosis.
  In patients who are unresponsive to initial steroid therapy, we stop using steroids through tapering. We observe the disease’s progress over time rather than continuing therapy with third- line drugs.

Bradycardia, VT and sudden death
In patients with advanced conduction disturbance, permanent pacemaker insertion should be considered according to general guideline indications. Not infrequently, conduction distur- bance can be recovered with steroid therapy. Still, it is generally believed that this recovery is transient, and conduction recovery does not invalidate the indication for pacemaker insertion. However, transient does not mean days or weeks; therefore, if the patient’s haemodynamic condition is stable in the presence of conduction disturbance, pacemaker insertion might be safely postponed if the conduction disturbance has recovered.
   Due to sudden cardiac death (SCD) in CS, inserting ICD should be considered when permanent pacing is indicated, as ICD has a pacemaker function. In the 2017 American Heart Association/American College of Cardiology/HRS guideline, ICD implantation is recommended in patients with evidence of extensive myocardial scarring by CMR or PET scan, even with an LVEF of >35%.³⁸   In one study analysing myocardial inflammatory diseases in the Finland study group registry,³⁹ the 5-year incidence of SCD in patients with AV block with an EF of <30% and VT was 34%. Even in patients with a lone AV block with normal EF and no VT, the 5-year-incidence of SCD was 9%. Based on this result, the authors recommended implanting ICD whenever permanent pacing is indicated. However, it is noteworthy that even though ICDs were implanted in 75% of the patients in patients with AV block with EF of <30% and VT, the 5-year incidence of SCD was 34%.
  ICD does not prevent non-arrhythmic causes of SCD, such as pump failure; therefore, if the patient experiences frequent shocks associated with VT, the adverse effect of ICD on LV func- tion should also be considered.
  VT is usually treated with antiarrhythmic agents. It is diffi- cult to draw conclusions regarding the effectiveness of steroids in suppressing VT. In one report,⁴⁰ corticosteroid therapy did not reduce the number of PVC or nonsustained VT (NSVT). However, a significant reduction in PVCs and NSVT was observed in patients with less advanced LV dysfunction (EF≥35%). Contrary to the expected positive effect, an increased ventricular arrhythmia burden with corticosteroid treatment has also been reported.41 However, the effects of individual immu- nosuppressive agents, such as third-line drugs, might not be the same as the effect of steroids on ventricular arrhythmia.⁴²
  Ablation therapy may be an option for patient refractory to medical therapy. However, despite successive ablation, the recur- rence rates are reported to be high.^{43 44} Several electrophysiolog- ical explanations for this poor outcome have been suggested,⁴⁵ and obstacles in catheter ablation do not seem to be solved easily.
   ICD insertion is one of the most challenging procedures in CS. Studies on risk stratification in patients with CS have been carried out using various modalities, such as the presence of DEs in CMR, perfusion defects and abnormal FDG uptake, induc- ible sustained VT in programmed electrical stimulation and RV involvement.⁴⁶ However, it is still too early to depend solely on one or more of these modalities for decision making on ICD implantation; therefore, following HRS guideline¹⁵ has been recommended (online supplemental table 3).

Correction notice This article has been corrected since it was first published to correct author name Jun-Bean Park.
Contributors DW-S wrote the manuscript and prepared the illustrations. JB-P reviewed the manuscript and supplemented the missed points to be dealt with in the manuscript.
ORCID iD Dae-Won Sohn http://orcid.org/0000-0002-1092-3285
http://heart.bmj.com/
http://dx.doi.org/10.1136/heartjnl-2022- 321379