Cardiovascular magnetic resonance native T₂ and T₂^* quantitative values for cardiomyopathies and heart transplantations: a systematic review and meta-analysis

Ventricular dysfunction in ischemic cardiomyopathies is triggered by impaired coronary blood supply to the myocardium [1]. In non-ischemic cardiomyopathy (NICM) many factors contribute to heart failure (HF) including hypertrophic cardiomyopathy (HCM), dilated cardiomyopathy (DCM) and restrictive cardiomyopathy [2, 3]. The prevalence of HF has been rising since the year 2000 and is shown to be related to the current lifestyle in Western Society [4, 5], with increasing populations with high cardiovascular risk (obesity, hypertension and type 2 diabetes mellitus (T2DM)) [6].

Early diagnosis of cardiomyopathy is important to initiate appropriate treatment [7, 8]. Physical examination and medical history are used to assess the probability of HF, however these assessments are non-specific in early diagnosis and therefore require additional tests [8, 9]. Electrocardiography (ECG) is also used in the first assessment of HF, and although an abnormal ECG increases the probability of HF, it has low specificity and provides little information to distinguish between cardiac diseases [8]. Transthoracic echocardiography was suggested as primary imaging test in the diagnostic pathway of HF because of its wide availability and low costs, and its cardiac function assessment enables fast decision making [8, 10], it however has limitations in distinguishing between underlying diseases [11]. Cardiovascular magnetic resonance (CMR) is the golden standard to detect cardiac remodelling by assessing the global cardiac function, it allows for regional function assessment with strain analysis and furthermore enables the assessment of myocardial fibrosis with late gadolinium enhancement (LGE) [8, 12‐14], whereas computed tomography is recommended to either exclude or to diagnose coronary artery disease [8]. Nevertheless, early myocardial structural changes are often qualitatively indistinguishable, and difficult to differentiate from overlapping findings in patients with high cardiovascular risk such as obesity, hypertension and T2DM [15‐18]. Consequently, misinterpretation of cardiac remodeling in these high cardiovascular risk groups may result in incorrect diagnosis and mistreatment. The changes occurring in cardiomyopathies, however, may affect myocardial tissue properties, which can be measured quantitatively by T₁, T₂ and T₂^* mapping as part of the CMR exam [19]. In line with this, the European Society of Cardiology recently described a shifting standards from the assessment of LGE towards the use of T₁ and T₂ mapping in their latest position statement [20]. The clinical utility of T₁ mapping has already been acknowledged and included in some guidelines [8, 13, 21, 22]. In addition, other guidelines also advocate to include T₂ and T₂^* mapping instead of T₂-weighted imaging [20, 22‐24].

The Society for Cardiovascular Magnetic Resonance (SCMR) released clinical recommendations about parametric imaging in CMR [22]. T₂ mapping values vary due to different water concentrations in the myocardium and therefore T₂ mapping could be useful in infiltrative cardiomyopathies, such as iron overload and Anderson-Fabry disease, and in myocardial injury diseases featuring edema, necrosis, and hemorrhage formation [22, 25, 26]. Furthermore, T₂ could contribute in the diagnosis of heart transplant rejections as edema correlates with acute heart transplant rejection [22, 27]. In addition to T₂, T₂^* mapping values mainly depend on magnetic field inhomogeneities and are therefore clinically useful in iron related diseases, and also enable assessment of hemorrhage formation [22, 28, 29].

Reference values of T₂ and T₂^* mapping in healthy subjects have been investigated in multiple studies [30‐33]. The heterogeneity of the data caused by different field strengths, imaging techniques and settings underlines the need for local reference values [22, 33]. The objective of this study was to perform a meta-analysis to determine the weighted mean of myocardial T₂ and T₂^* mapping values in the HF-related cardiomyopathies and heart transplantations, and standardized mean differences (SMD) with healthy controls. Knowledge of these ranges can help determine the clinically applicability of quantitative techniques. Furthermore, we aim to investigate the presumed heterogeneity of studies leading to variation in mapping outcomes, to emphasize the importance of mapping standardization.

Quantitative analysis of factors that modulate myocardial T₂ and T₂^*, such as edema, lipids and paramagnetic iron-containing depositions, can potentially provide additional diagnostic information to distinguish between myocardial diseases and healthy myocardium. This meta-analysis confirmed that T₂ mapping can help differentiate between healthy subjects and patients affected by MI, DCM, myocarditis or heart transplantation, since T₂ values were higher in these populations [22]. Although T₂ mapping has been expected to be sensitive to iron as well [22], no significantly lower T₂ values were found between iron overload related diseases and healthy myocardium (P = 0.30). On sarcoidosis, SLE, amyloidosis, sarcoidosis, Anderson-Fabry disease and HCM insufficient studies were reported for further analysis, nevertheless the available data suggested T₂ values to be higher within these diseases, with an exception for Anderson-Fabry disease patients without LVH. Furthermore, this meta-analysis confirmed that T₂^* mapping can differentiate between healthy myocardium and myocardium affected in MI and iron overload, since T₂^* values were lower in both of these populations [22]. For HCM, DCM and hypertension patients, the limited available T₂^* mapping studies also gave some indication of lower T₂^* values compared to controls, however this was overall not significant. For all included cardiac diseases in this meta-analysis the T₂ values were higher, with iron overload patients as an exception showing lower T₂ values, and T₂^* values were lower. These similarities in T₂ and T₂^* values between cardiac diseases prevent further differentiation in disease type, as opposed to differentiation from the healthy.

Reported T₂ and T₂^* values in healthy subjects showed large variation between studies, which could partly be due to the lack of acquisition standardization. In the standardized CMR imaging guideline and protocol published in 2013 [194], T₂^* mapping was only described as a clinical applicable technique to assess cardiac iron deposition and T₂ mapping was defined as a research-domain technique [194, 195]. T₂ mapping sequences were stated as optional since there was no standardization yet [194], which led to different acquisition approaches and therefore potentially acquisition related variation in T₂ values. In 2017, clinical recommendations were released regarding parametric imaging of both T₂ and T₂^* mapping and defined standardized data acquisition and analysis [22]. They stated that local healthy T₂ and T₂^* values should be determined in order to clinically use these quantitative techniques, which is now confirmed by this meta-analysis considering the wide variation of healthy T₂ and T₂^* values (Figs. 2, 3, 5 and 6). The use of normal scan results of clinically referred patients could be used to determine reference values, however this is not recommended due to referral bias. Age- and gender-matching of the control group is necessary [22], since both are known to influence T₂ and T₂^* values [30]. Furthermore, the clinical recommendations also stated specific imaging protocols, technical requirements of sequences and image planning for T₂ and T₂^* mapping, which should reduce variability in image acquisition from then onward [22]. This meta-analysis includes multiple studies that were published prior to this guideline and showed the heterogeneity to be significantly influenced by the sequence based covariates, which has previously already been concluded from a direct comparison between sequences [196]. This analysis also showed the variation between CMR vendors with on 1.5 T healthy control T₂ values of 54.9 ± 3.3 ms at Philips (n = 13 studies) and 50.0 ± 2.5 ms at Siemens (n = 22) and T₂^* values of 34.1 ± 6.5 ms at Philips (n = 5), 30.8 ± 4.5 ms at Siemens (n = 3) and 55.0 ± 13.0 ms at General Eletric (GE) (n = 1), and on 3 T healthy control T₂ values of 44.7 ± 5.8 ms at Philips (n = 6) and 48.0 ± 3.0 ms at Siemens (n = 5), and T₂^* values of 23.9 ± 4.7 at Philips (n = 2), 21.0 ± 4.8 ms at Siemens (n = 1) and 21.0 ± 6.4 ms at GE (n = 1). These differences in vendor and field strength should be kept in mind when T₂ and T₂^* values are used within a clinical protocol.

In addition to the clinical guideline on T₂ and T₂^* acquisitions [22], following the recommendations in image analysis could reduce the non-physiological variation of T₂ and T₂^* values. The clinical recommendations on acquisition and ROI placement are described specifically per disease [22], and this meta-analysis confirmed the different approaches in analysis. In general the ROI should be placed outside positive LGE myocardium areas and include non-fibrous myocardium [22]. T₂ values measured in positive LGE myocardium should therefore be interpreted cautiously. Analysis of T₂ in diffuse diseases, such as HCM and DCM, were mostly performed based on one or three short axis (SAx) slices using global assessment [162, 164‐169], as recommended [22]. In patchy diseases, such as amyloidosis and Anderson-Fabry disease, the recommendations state that the T₂ analysis should also include a single 3 chamber or 4 chamber view acquisition additionally to basal and mid-ventricular SAx slices [22]. Only one study actually followed these recommendations [158], while for the other cardiac patchy disease studies one or more recommended slices were not included [155‐157]. In focal diseases, such as MI and myocarditis, the ROI differs between patients because the location of the abnormality is different, and therefore the guideline recommends multiple SAx acquisition to cover the whole myocardium and to place the ROI in visually abnormal myocardium [22]. Most included studies in this meta-analysis therefore acquired multiple SAx slices [51, 54‐56, 61, 63, 65], however some studies acquired only one [60] or three [49] SAx slices at the level of the infarcted area, which is more prone to missing the infarct core. In the studies with myocarditis patients mapping acquisition was generally also performed over multiple SAx covering the whole myocardium [38, 171‐173, 175‐180, 182, 183, 185], however in some studies the T₂ values were only acquired from a LGE hyperintense based ROI [25, 174, 181, 184, 187]. Also, studies including MI, often distinguish between the infarct region or core and use remote myocardium as the healthy control tissue. In these studies the ROI placement was generally based on LGE hyperintense regions [26, 41, 49, 51, 57, 58, 60‐63, 65, 67, 68], 2SD change of T₂ signal intensity [40, 43, 54, 56, 59, 60] or T₂^* values [41, 43, 56]. This meta-analysis showed that ROI placement significantly influences the T₂ and T₂^* outcome and the separate analysis showed the infarct zone to have a larger T₂ difference with controls than the infarct core, while the infarct core showed a larger T₂^* difference with controls than the infarct zone. Lastly, for studies including iron overload patients most T₂^* measurements were performed in the intraventricular septum for reproducibility, because the lateral wall often contains dephasing artefacts. Nevertheless, some studies reported an average of the mid-ventricular SAx slice [87, 115, 119, 134] or the entire myocardium [106, 125, 127‐132], which especially on 3 T [127] could lead to some unrealistic T₂^* values due to aforementioned artefacts.

In this meta-analysis including MI patients other covariates aside from the ROI placement had a significant effect on T₂ and T₂^* mapping outcomes. These covariates included the use of remote myocardium as control values instead of healthy controls, the timing of CMR acquisition after reperfusion, and the sequence that was used. The first covariate that included the use of remote myocardium as control, showed that remote myocardium is physiologically different from healthy tissue and therefore is not an appropriate control tissue [197, 198]. Followed by the second covariate for timing of the CMR imaging after PCI, for which histologically is verified in swine that edema and haemorrhage formation peaks in the acute phase 2 h and 7 days post-PCI [199]. These peaks were also detected in the acquired T₂ values in humans at the same day and at 10 days post-PCI, compared to 3 days post-PCI [43]. These results were contradicted by another study [64] that reported higher T₂ values at 3 days post-PCI compared to the same day or at 7 days post-PCI. The third covariate showed that the use of a spin-echo based sequence provides larger differences between MI patients and controls, than the gradient-echo-spin-echo or T₂-prepared balanced steady-state free procession sequences, while the latter two are currently recommended in the general guideline [22]. Lastly due to the remaining high heterogeneity of the MI meta-analysis other covariates are expected to influence the T₂ and T₂^* mapping outcomes in addition to the ones identified here.

In this meta-analysis including heart transplant patients the main distinct covariate was the rejection status of the transplanted heart. Acute cellular rejection is characterized by infiltration of inflammatory cells accompanied with edema resulting in higher T₂ values [22, 200], which was also reported in most included studies [22, 27, 71‐73, 75, 76, 200]. Nevertheless, patients with negative biopsies also showed higher T₂ values than controls [69, 71, 75], suggesting that the higher T₂ values in heart transplant patients may also be related to the inflammatory changes from the transplantation process. The exploratory meta-analysis, however, indicated that positive rejection was a significant covariate to result in larger differences of T₂ values between heart transplant patients and healthy controls [27, 72, 73, 77], and therefore further research is needed to investigate the clinical applicability of T₂ mapping for early detection of heart transplant rejection.

In this meta-analysis all transfusion-dependent diseases leading to iron overload were evaluated in one group including thalassemia, sickle cell disease and anaemias [201]. The overall average T₂^* value for iron overload patients was 27.2 ± 13.7 ms, which was above the established iron overload cut-off (T₂^* < 20 ms) [195]. This could be due to the fact that most studies reported T₂^* values without distinguishing between cardiac or non-cardiac iron overload involvement. Some studies provided T₂^* values of cardiac involved patients using < 20 ms as a clinical cut-off [22]. Consequently, the mean T₂^* value of these cardiac involved patients was only 11.8 ± 3.7 ms, which was significantly lower than the controls. The type of controls should ideally only include healthy volunteers, however in some studies also non-cardiac involved iron overload patients were used as controls. The T₂^* value from real healthy volunteers of 32.4 ± 5.6 ms [79, 81, 85, 88, 93, 107, 118, 133] was lower than the 35.7 ± 6.4 ms from non-cardiac iron overload patients [95, 96, 104, 113, 114, 124, 127, 132], and therefore the accuracy of the T₂^* < 20 ms cut-off to establish cardiac involvement could be challenged. The current recommendation advises to perform T₂^* mapping on 1.5 T, since higher field strengths show more susceptibility artefacts [22]. Nonetheless, two studies [81, 88] were performed at 3 T as well as 1.5 T including patients and controls, in which ROI placement was performed at the mid-ventricular septum to avoid susceptibility artefacts [22]. As expected, these studies showed a larger SMD between healthy controls and iron overload patients at 3 T compared to 1.5 T (SMD of − 0.27 and − 0.16), since the transverse relativity of paramagnetic substrates increases with field strength [202]. These last findings show that iron overload evaluation on 3 T seems to be a trade-off between increased risk on artefacts and a higher iron sensitivity.

Furthermore, T₂ mapping was expected to be sensitive for iron overload [22], however this was not unequivocally confirmed by this meta-analysis (SMD = − 0.54, P = 0.30). One study performed on 1.5 T and 3 T showed no statistically significant T₂ changes in iron overload patients [81], while others did show clear changes in T₂ values [82, 93, 101]. In this first study only 6% of their patients had cardiac involvement, which might explain the lack of change in T₂. The other studies showed a high correlation between T₂ and T₂^* changes and significantly lower T₂ values in patients with cardiac involved iron overload compared to healthy controls suggesting that T₂ to could indeed be sensitive to iron overload [82, 93, 101]. More research is needed to validate this conclusion.

In Anderson-Fabry disease only patients with LVH showed significantly higher T₂ values compared to healthy controls [158]. Previous research showed that native T₁ mapping is the most sensitive CMR parameter in Anderson-Fabry disease and that Anderson-Fabry disease patients showed lower T₁ values than controls regardless of LV function and morphology, and therefore T₁ mapping is also sensitive to distinguish between controls and Anderson-Fabry disease patients without LVH [203]. One study, which was not included within this meta-analysis because it was published previous to our search period, also reported higher T₂ values in Anderson-Fabry disease patients compared to both HCM patients and healthy controls, suggesting that T₂ mapping is also a sensitive CMR marker to early assess cardiac involvement in Anderson-Fabry disease patients without LVH [204].

The higher T₂ values in DCM patients found in this meta-analysis confirmed the immunohistologal evidence of chronic myocardial inflammation for this disease [205]. Studies reporting T₂ values of DCM subgroups seemed contradicting, since one study [166] showed higher T₂ values in severe DCM compared to mild DCM (P <  0.05), while another [167], though not significant, showed lower T₂ values in severe DCM compared to mild DCM. Nevertheless, overall higher T₂ values in DCM patients was confirmed by this meta-analysis.

This meta-analysis including studies with myocarditis patients confirmed the expected higher T₂ values in the acute phase. All studies reported significantly higher T₂ values except for one study that showed non-significantly higher T₂ values in the acute phase compared to healthy controls, with 65.3 ± 45.4 ms and 53.7 ± 31.0 ms, respectively, which was mainly due to the broad SD of both groups [184]. Aside from the higher T₂ values in the acute phase, a follow-up study showed that 3 and 12 months after symptom onset the T₂ values returned to normal [174]. Another follow-up study confirmed these normal T₂ values at 189 days after symptom onset, and also showed that after 40 days the T₂ values were still significantly higher compared to healthy controls, with 52.4 ± 1.0 ms and 50.4 ± 2.3 ms, respectively [185]. These follow-up studies suggest that T₂ mapping in myocarditis is most valuable in the acute phase in addition to the Lake Louise criteria that include histology and CMR with T₁- and T₂-weighted imaging.

The single study that reported T₂ values from HCM patients and controls showed significantly higher T₂ values in patients [158]. Two studies compared the T₂^* values from HCM patients with healthy controls, however their results were contradicting. One study at 1.5 T reported significantly lower T₂^* values in HCM patients compared to controls with 26.2 ± 4.6 ms and 31.3 ± 4.3 ms, respectively [159], whereas the other study at 3 T reported no significant difference with 22.3 ± 4.1 ms and 21.0 ± 6.4 ms, respectively [160]. Since early treatment is key for HCM patients, it is important to be able to distinguish LVH changes due to either HCM or to hypertension. Differentiating between HCM and hypertension related LVH using only parametric imaging is not possible, as this differentiation depends on multiple clinical factors [13]. Nevertheless one study reported on hypertension patients and showed lower T₂^* values at 3 T for both hypertension patients with LVH (23.8 ± 3.1 ms) and without LVH (28.6 ± 4.2 ms) compared to healthy controls (30.8 ± 2.7 ms) [50]. Based on these limited available studies no conclusion can be drawn on the clinical relevance of T₂ and T₂^* mapping. More research could enable to determine the clinical applicability of these mapping techniques, while T₁ mapping has already shown to be promising in distinguishing hypertension related LVH and HCM [21, 206]. Furthermore, as the incidence of cardiomyopathies is related to obesity and T2DM [8] it is important to determine whether these high cardiovascular risk factors cause myocardial tissue adaptation and if these are distinguishable with quantitative techniques. Unfortunately, no T₂ and T₂^* mapping of these risk populations is yet, and therefore we have to rely on the values of cardiac diseases without considering these risk factors.

This meta-analysis showed that T₂ and T₂^* values of both patients and healthy controls demonstrate variation between studies related to differences in population demographics, CMR vendor, acquisition methods and analysis approach. This variation limits comparison between centers and therefore each center requires local T₂ and T₂^* reference values to distinguish affected myocardium in cardiomyopathies from healthy myocardium. To this end reference values should be obtained in, preferably matched, healthy controls using the same CMR acquisition method as in patient care. Although similarities of changes in T₂ and T₂^* values between cardiac diseases limits direct differentiation, this paper provides T₂ and T₂^* mapping data which, together with other CMR parameters such as T₁ mapping, ECV and LGE, can help to differentiate between cardiac disease entities.