Background

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, 12,13,14], 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 [15,16,17,18]. 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, 22,23,24].

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 [30,31,32,33]. 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.

Materials and methods

Search strategy

The study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [34] and the Cochrane Handbook for Systematic Review [35]. Three independent investigators (GS, MvdB and LH) systematically searched for eligible studies published between January 2011 and September 2019 in PubMed/MEDLINE and Embase applying CMR T2 or T2* mapping in humans. The search contained terms related to T2 or T2* mapping and cardiac diseases (full search terms are listed in Supplementary Data 1).

In this meta-analysis we accepted published results from randomized control trials, cohort studies and observational studies in peer-reviewed journals if they included adults aged 18 years and older with NICM or ischemic cardiomyopathy, heart transplant patients or adults with increased cardiovascular risk, and reported CMR derived T2 and/or T2* mapping values acquired at 1.5 T or 3 T. Studies were excluded if the article was not available in English or in full text.

Study selection

Titles and abstracts proposed by the databases were assessed for eligibility by one author and checked by a second author (GS, MvdB and LH). After consensus between these investigators, the full-text reports of these eligible studies were independently assessed by two investigators for final inclusion. The study quality was systematically evaluated with the Newcastle-Ottawa quality assessment scale (NOS) [36]. This scale evaluated the study quality on three domains: selection and definition of included populations (0–4 points); comparability of the controls (0–2 points); and ascertainment of the outcome (0–3 points).

Data collection

Data were extracted from the included studies by one author and checked by a second author (GS, MvdB and LH). Relevant data regarding patient characteristics, such as; study population, age, gender, body mass index, T2 and T2* values, as well as CMR imaging acquisition related information, such as; field strength, vendor, sequence and sequence parameters were extracted. Data were reported as mean ± standard deviation (SD) and data reported as median with interquartile or full range were converted using the methodology of Hozo et al. [37]. Healthy control data were extracted if available.

Data analysis

The included data were divided into two groups of reported T2 and T2* values per disease and combined into a random effects model to determine the SMD and the 95% confidence interval (CI). The heterogeneity of the included studies was defined with I2 being significant if I2 ≥ 50% (P <  0.05) by using a χ2 test. This heterogeneity was further tested by a meta-regression, sensitivity and bias analysis. Available covariates were tested for their association with the myocardial T2 and T2* values using a backwards elimination model and remaining significant covariates (P <  0.05) were included into a mixed effect model of the data. Publication bias was assessed by inspection of the funnel plots with the Egger regression asymmetry test and a sensitivity analysis was performed by omitting each study sequentially and recalculating the model. A meta-analysis was performed in each population with at least 10 published studies, as stated by the PRISMA guideline [34]. Review Manager (RevMan) v. 5.3 (Cochrane Collaboration, Copenhagen, Denmark) was used to determine the random effect models and the package “metaphor” in R v. 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria) was used for the mixed effect models, bias and sensitivity analysis.

Results

Literature search

The search in PubMed and Embase revealed respectively 555 and 545 articles, and one article was manually added [38]. After removal of the duplicates, 704 articles remained for evaluation of title and abstract which resulted in 154 articles included for the final meta-analysis (Table 1). In the final exclusion step based on full text assessment, we excluded studies which presumably included (mostly) the same patient population as other included studies based on authors and method; the study with the least inclusions was excluded. The PRISMA flow diagram with rationale for exclusion is provided in Fig. 1. The number of studies per population was described as total studies (number of studies reporting T2 data & number of studies reporting T2* data): A total of 31 (22 T2 & 13 T2*) studies were included in the myocardial infarction (MI) population [26, 39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68], 11 (11 T2 & 0 T2*) in heart transplantation [27, 69,70,71,72,73,74,75,76,77,78], 70 (5 T2 & 70 T2*) in iron overload [79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148], 2 (2 T2 & 0 T2*) in sarcoidosis [149, 150], 4 (4 T2 & 0 T2*) in systemic lupus erythematosus (SLE) [151,152,153,154], 2 (2 T2 & 0 T2*) in amyloidosis [155, 156], 2 (2 T2 & 0 T2*) in Anderson-Fabry disease [157, 158], 4 (2 T2 & 2 T2*) in HCM [159,160,161,162], 9 (7 T2 & 2 T2*) in DCM [160, 163,164,165,166,167,168,169,170], 19 (19 T2 & 0 T2*) in myocarditis [25, 38, 171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187] and 1 (0 T2 & 1 T2*) in hypertension [188] (Table 1). The absolute T2 and T2* values are dependent on field strength [189, 190], therefore the average mapping values were noted separately for 1.5 T and 3 T, and it was also used as covariate in the meta-regression analysis. T2 and T2* mapping obtained in control subjects were recorded as values from healthy subjects, unless the control population was explicitly defined otherwise in the “population” column of Table 1.

Fig. 1
figure 1

Overview of the study review process according to the PRISMA flow diagram

Full size image
Table 1 Characteristics of the included studies in the meta-analysis
Full size table

Study quality

None of the included studies received the maximum NOS quality score (Table 1). All studies without healthy controls automatically received limited scores in the matching and selection section. Only 57 of the 154 included studies reported control values of healthy subjects. The case definition of patients and the ascertainment of mapping values were adequate in all studies.

Myocardial infarction

The weighted mean T2* values at 1.5 T in myocardial infarction (MI) patients was 28.5 ± 6.8 ms and 34.7 ± 3.7 ms in controls [39,40,41,42,43,44,45,46,47,48,49] (Table 1, Fig. 2). At 3 T, these were 22.0 ± 3.7 ms in MI patients and 29.6 ± 2.7 ms in controls [50, 51] (Table 1, Fig. 3). The meta-analysis confirmed significantly lower T2* values in MI patients (SMD = − 1.99, 95% Cl [− 2.70, − 1.27], P <  0.01, I2 = 98%, Fig. 4). Most studies performed CMR in ST-elevation myocardial infarction (STEMI) patients post percutaneous coronary intervention (PCI) in the acute phase [39,40,41,42,43,44, 46,47,48,49,50,51]. Some studies performed follow-up in these patient groups [42,43,44, 47, 49, 50] and Mohammadzadeh et al. [45] was the only study including non-STEMI (NSTEMI) patients. Most studies reported T2* values of multiple regions-of-interest (ROI) in the myocardium (Table 1). Although none of the tested covariates was significant, the difference in T2* values seemed larger in the infarct cores compared to the infarct zone as a whole. Significant funnel asymmetry was found for the random effects model suggesting eight missing studies with negative results (P <  0.01), while the mixed effects model did not show funnel asymmetry (P = 0.60).

Fig. 2
figure 2

Weighted mean T2* values and weighted standard deviations (SD) of all included papers reporting T2* values of both patients (black squares) and controls (grey squares) measured at 1.5 T. The number of included patient (p) and control (c) measurements for each population is reported above the graph. MI myocardial infarction, IO iron overload, HCM hypertrophic cardiomyopathy, DCM dilated cardiomyopathy, HTN hypertension

Full size image
Fig. 3
figure 3

Weighted mean T2* values and weighted standard deviations (SD) of all included papers reporting T2* values of both patients (black squares) and controls (grey squares) measured at 3 T. The number of included patient (p) and control (c) measurements for each population is reported above the graph. MI myocardial infarction, IO iron overload, HCM hypertrophic cardiomyopathy, DCM dilated cardiomyopathy, HTN hypertension

Full size image
Fig. 4
figure 4

Standardized mean differences between T2* of myocardial infarction (MI) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

The heterogeneity was not corrected with the existing covariates and therefore a second analysis was performed where the reported T2* values were divided in infarct zone or infarct core groups. The infarct zone, which is determined by LGE, is the affected myocardium characterized by edema excluding the hypo-intense core, which is the center in the infarct zone with T2* values < 20 ms identifying the presence of hemorrhage [40, 50]. Although during myocardial infarction no haemorrhagic core is present, the patients were referred for CMR after PCI in most studies. The process of reperfusion after PCI frequently leads to simultaneous microvascular obstruction and intramyocardial haemorrhage within the infarct zone [41, 191].

Eight studies [39,40,41, 43,44,45, 48, 50] explicitly reported infarct zone values. The weighted mean T2* value at 1.5 T of the infarct zones was 32.3 ± 5.4 ms and at 3 T this was 22.4 ± 2.8 ms (Fig. 1, Supplementary Data 2). These T2* values also resulted in significantly lower values compared to controls (SMD = − 1.21, 95% Cl [− 1.83, − 0.59], P <  0.01, I2 = 95%), and with a significant heterogeneity. Furthermore, infarct core values were explicitly reported in five studies [40, 41, 43, 46, 51]. The weighted mean T2* value at 1.5 T of infarct cores was 16.1 ± 4.2 ms and at 3 T this was 16.1 ± 7.6 ms (Fig. 1, Supplementary Data 2). These infarct core values showed a larger SMD (SMD = − 4.00, 95% Cl [− 5.67, − 2.32], P <  0.01, I2 = 98%), while the heterogeneity remained significant. Multiple studies reported the remote myocardium as control which had a weighted mean T2* value at 1.5 T of 34.0 ± 4.9 ms and 30.5 ± 1.0 ms at 3 T (Fig. 1, Supplementary Data 2).

The weighted mean T2 values at 1.5 T in MI patients was 58.5 ± 5.8 ms and 49.3 ± 2.6 ms in controls [26, 40, 41, 43, 49, 52,53,54,55,56,57,58,59,60,61,62,63] (Table 1, Fig. 5). At 3 T, these were 60.3 ± 9.7 ms in MI patients and 44.0 ± 3.8 ms in controls [51, 64,65,66,67,68] (Table 1, Fig. 6). Most studies restricted their inclusion to STEMI patients [40, 41, 43, 49, 51, 54,55,56,57,58,59,60, 63,64,65, 68], however some studies included specifically NSTEMI patients [52, 62, 67] and others included both STEMI and NSTEMI patients [26, 53, 61, 66]. Besides two studies [52, 62], patients in all studies underwent CMR post-PCI in the acute phase and a few studies also included follow-up data [40, 42, 43, 49, 53, 64]. T2 values of different ROIs in the myocardium were reported (Table 1), nevertheless all studies showed higher T2 values in all ROIs of MI patients except for studies reporting values of the hemorrhagic core [40, 41]. The meta-analysis confirmed significantly higher T2 values in MI patients (SMD = 2.17, 95% CI [1.79, 2.54], P <  0.01, I2 = 96%, Fig. 7). The age and percentage of men in the control group, the time between intervention and the CMR, the field strength, the type of control (remote myocardium versus healthy controls), the type of CMR acquisition sequence, the ROI location and the left ventricular ejection fraction (LVEF) in patients were significant covariates. There were no other significant residual factors remaining that accounted for the high remaining heterogeneity (I2 = 78%), though there are probably other covariates which were not tested due to insufficient data. Publication bias was detected with five possibly missing studies, however no significant asymmetry was found for either the random effects model (P = 0.10) or the mixed effects model (P = 0.55).

Fig. 5
figure 5

Weighted mean T2 values and weighted standard deviations (SD) of all included papers reporting T2 values of both patients (black squares) and controls (grey squares) measured at 1.5 T. The number of included patient (p) and control (c) measurements for each population is reported above the graph. MI myocardial infarction, Trans heart transplant, IO iron overload, SA sarcoidosis, SLE systemic lupus erythematosus, AM amyloidosis, HCM hypertrophic cardiomyopathy, DCM dilated cardiomyopathy, MC myocarditis

Full size image
Fig. 6
figure 6

Weighted mean T2 values and weighted standard deviations (SD) of all included papers reporting T2 values of both patients (black squares) and controls (grey squares) measured at 3 T. The number of included patient (p) and control (c) measurements for each population is reported above the graph. MI myocardial infarction, Trans heart transplant, IO iron overload, SA sarcoidosis, SLE systemic lupus erythematosus, AM amyloidosis, HCM hypertrophic cardiomyopathy, DCM dilated cardiomyopathy, MC myocarditis

Full size image
Fig. 7
figure 7

Standardized mean differences between T2 of myocardial infarction (MI) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

The ROI location was one of the covariates and therefore an additional analysis was performed where the reported T2 values were divided in infarct zone and infarct core groups. Infarct zone T2 values were reported in 18 studies [26, 40, 43, 51, 53, 54, 56,57,58, 60,61,62,63,64,65,66,67,68]. The weighted mean T2 value at 1.5 T of infarct zones was 63.7 ± 6.4 ms and at 3 T this was 63.5 ± 10.5 ms (Fig. 2, Supplementary Data 2). The difference between patients and controls was larger when considering only the infarct zone values (SMD = 2.63, 95% Cl [2.25, 3.01], P <  0.01, I2 = 93%). The meta-analysis showed older patients, a short period between intervention and CMR, lower LVEF in patients and performing CMR on 1.5 T to increase the difference with controls. The used CMR acquisition sequence was also found as significant covariate, nevertheless none of the specified sequences provided clearly larger differences. There were no other significant residual factors remaining that accounted for the heterogeneity (I2 = 80%). Again, publication bias was found with two missing studies, however no significant asymmetry was found for either the random effects model (P = 0.76) or the mixed effects model (P = 0.58). Core T2 values were reported in five studies [40, 41, 43, 56, 60]. The weighted mean T2 value at 1.5 T of infarct cores was 51.9 ± 4.6 ms and at 3 T no values were reported (Fig. 2, Supplementary Data 2). Including only the T2 values of the infarct cores resulted in a smaller difference between patients and controls (SMD = 0.83, 95% Cl [0.37, 2.44], P <  0.01, I2 = 91%). The weighted mean T2 value at 1.5 T of remote myocardium was 49.2 ± 2.5 ms and at 3 T this was 45.0 ± 3.0 ms (Fig. 2, Supplementary Data 2).

Heart transplant

The weighted mean T2 values at 1.5 T in heart transplant patients was 54.6 ± 5.2 ms and 49.2 ± 2.5 ms in controls [27, 69,70,71,72,73,74,75,76,77,78] (Table 1, Fig. 5). All studies showed higher T2 values in patients compared to controls, only for all subgroups including patients with positive rejection biopsy these values were significantly higher. This meta-analysis confirmed significantly higher T2 values in the myocardium of heart transplant patients (SMD = 1.05, 95% CI [0.69, 1.41], P <  0.01, I2 = 65%, Fig. 8). An exploratory meta-regression analysis indicated that the rejection status, the LVEF and patient age caused the heterogeneity without remaining significant residual factors (I2 = 1%). Transplant rejection, lower LVEF and older patients resulted in larger differences between patients and controls.

Fig. 8
figure 8

Standardized mean differences between T2 of heart transplant patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

The cardiac transplant rejection was a significant covariate and therefore the population was divided between positive and negative rejection biopsies. The weighted mean T2 values in patients with a positive biopsy [27, 69, 71, 73,74,75] was 56.4 ± 3.3 ms and 52.5 ± 3.9 ms in patients with a negative biopsy [27, 69, 71,72,73,74,75,76] (Fig. 2, Supplementary Data 2). None of the studies to heart transplantation described T2 values acquired at 3 T or reported T2* values.

Iron overload

The weighted mean T2* values at 1.5 T in iron overload patients was 27.2 ± 13.7 ms and 36.1 ± 6.3 ms in controls [79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147] (Table 1, Fig. 2). At 3 T, these were 21.8 ± 7.8 ms in iron overload patients and 22.4 ± 3.8 ms in controls [81, 88, 127, 148] (Table 1, Fig. 3). The meta-analysis confirmed significantly lower T2* values in iron overload patients (SMD = − 2.39, 95% CI [− 3.28, − 1.49], P <  0.01, I2 = 98%, Fig. 9). The patient populations contained iron overload patients with both cardiac involvement (T2* < 20 ms) and without cardiac involvement (T2* ≥ 20 ms). Each study that included both iron overload patients and controls showed significantly lower T2* values in patients [85, 93, 95, 96, 104, 107, 113, 114, 118, 124, 128, 132, 133, 136, 139], except for two studies that showed non-significant lower T2* values [81, 88] and one study that showed non-significantly higher T2* values in patients compared to controls [79]. The type of control was found as a covariate which meant using non-cardiac involved iron overload subjects as controls caused larger differences with patients than using healthy controls. The type of patients was also found as covariate; using a population with proven cardiac involvement caused larger differences with controls than using a mix of non-cardiac and cardiac involved iron overload patients. Furthermore, the number of echoes used in the T2* sequence was determined as a covariate. These covariates, however, only partly accounted for the heterogeneity in the mixed effects model (I2 = 80%), while other tested regressors (age of patient and control population, percentage of men in patient and control population, CMR vendor, field strength and the serum ferritin concentration in patients) had no significant influence. Based on the high remaining heterogeneity there should be other covariates which were not tested due to insufficient data. Significant funnel asymmetry (P <  0.01) was only found for the random effects model suggesting five missing studies with populations showing higher T2* values compared to healthy subjects.

Fig. 9
figure 9

Standardized mean differences between T2* of iron overload patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

The type of iron overload patient was one of the covariates and therefore an additional analysis was performed on T2* values from cardiac involved iron overload patients (T2* < 20 ms) [93, 95, 96, 104, 113, 114, 123, 124, 128, 132, 136, 139, 143, 145]. The weighted mean T2* value at 1.5 T in cardiac involved iron overload patients was 11.8 ± 3.7 ms and at 3 T no T2* values were reported (Fig. 1, Supplementary Data 2). This analysis also showed significantly lower T2* values for cardiac involved iron overload patients compared to controls (SMD = − 3.59, 95% CI [− 4.69, − 2.48], P <  0.01, I2 = 97%) and this difference was also larger than controls compared to the overall iron overload population.

The weighted mean T2 values at 1.5 T in iron overload patients was 56.0 ± 13.6 ms and 58.3 ± 3.2 ms in controls [81, 82, 101] (Table 1, Fig. 5). At 3 T, these were 53.2 ± 6.2 ms in iron overload patients and 52.0 ± 5.5 ms in controls [81, 93] (Table 1, Fig. 6). Kritsaineeboon et al. [81] reported no significant changes in T2 values for iron overload patients at both 1.5 T and 3 T, while Camargo et al. [93] reported lower T2 values in iron overload patients at 1.5 T. The random effects models of all studies combined resulted in no significantly lower T2 values for iron overload patients compared to controls (SMD = − 0.54, 95% Cl [− 1.56, 0.48], P = 0.30, I2 = 86%, Fig. 10).

Fig. 10
figure 10

Standardized mean differences between T2 of iron overload patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Sarcoidosis

The weighted mean T2 values at 1.5 T in sarcoidosis patients was 52.3 ± 3.8 ms and 49.0 ± 1.6 ms in controls [149] (Table 1, Fig. 5). At 3 T, these were 54.0 ± 12.2 ms in sarcoidosis patients and 45.0 ± 10.8 ms in controls [150] (Table 1, Fig. 6). This suggested higher T2 values in sarcoidosis patients (SMD = 0.87, 95% CI [0.55, 1.20], P <  0.01, I2 = 0%, Fig. 11). Insufficient studies were available for further analysis regarding covariates and publication bias, and there was no data that described T2* values.

Fig. 11
figure 11

Standardized mean differences between T2 of sarcoidosis patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Systemic lupus erythematosus

The weighted mean T2 values at 1.5 T in systemic lupus erythematosus (SLE) patients was 55.7 ± 4.9 ms and 50.6 ± 3.3 ms in controls [151, 152] (Table 1, Fig. 5). At 3 T, these were 57.3 ± 8.6 ms in SLE patients and 44.4 ± 4.0 ms in controls [153, 154] (Table 1, Fig. 6). This suggested higher T2 values in SLE patients (SMD = 1.39, 95% CI [0.34, 2.44], P <  0.01, I2 = 93%, Fig. 12). Insufficient studies were available for further analysis regarding covariates and publication bias, and there were no data that described T2* values.

Fig. 12
figure 12

Standardized mean differences between T2 of systemic lupus erythematosus patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Amyloidosis

The weighted mean T2 values at 1.5 T in amyloidosis patients was 55.3 ± 4.2 ms and 50.2 ± 2.7 ms in controls [155, 156] (Table 1, Fig. 5). All included studies reported higher T2 values in amyloidosis patients (SMD = 1.62, 95% CI [1.19, 2.06], P <  0.01, I2 = 76%, Fig. 13). Although insufficient studies were available for further analysis regarding covariates and publication bias, both included studies reported higher T2 values in amyloid light-chain amyloidosis than in transthyretin amyloidosis. Furthermore, there were no studies performed with T2 values on 3 T and there was no data that described T2* values.

Fig. 13
figure 13

Standardized mean differences between T2 of amyloidosis (AM) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Anderson-Fabry disease

The weighted mean T2 value at 1.5 T in Anderson-Fabry disease patients was 57.7 ± 3.0 ms [157, 158] (Table 1, Fig. 5). One study reported T2 values in controls of 47.5 ± 2.4 ms [158], suggesting a trend to higher T2 values in Anderson-Fabry disease patients (SMD = 0.52, 95% Cl [− 0.23, 1.28], P = 0.17, I2 = 71%, Fig. 14). The higher T2 values in Anderson-Fabry disease patients were caused by the reported T2 values in Anderson-Fabry disease patients with left ventricular hypertrophy (LVH) (50.4 ± 3.8 ms), while patients without LVH showed similar T2 values (47.8 ± 1.7 ms) to controls. Insufficient studies were available for further analysis regarding covariates and publication bias. Furthermore, there were no studies performed with T2 values on 3 T and there were no data that described T2* values.

Fig. 14
figure 14

Standardized mean differences between T2 of Fabry disease patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Hypertrophic cardiomyopathy

The weighted mean T2* values at 1.5 T in HCM patients from one study was 26.4 ± 4.4 ms and 31.3 ± 4.3 ms in controls [159] (Table 1, Fig. 2). At 3 T, these were 22.3 ± 4.1 ms in HCM patients and 21.0 ± 6.4 ms in controls [160] (Table 1, Fig. 3). The study performed at 1.5 T reported values in subgroups based on the presence of fibrosis (with or without LGE) and in both subgroups the T2* value was lower compared to controls, which was only significant in the group with LGE presence [159]. In the study performed at 3 T there, however, was no significant difference in T2* values between HCM patients with or without LGE presence. As result, the analysis showed a no significant difference between HCM patients and controls (SMD = − 0.61, 95% CI [− 1.58, 0.36], P = 0.22, I2 = 87%, Fig. 15). Insufficient studies were available for further analysis regarding covariates and publication bias.

Fig. 15
figure 15

Standardized mean differences between T2* of hypertrophic cardiomyopathy (HCM) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

The weighted mean T2 value at 1.5 T in HCM patients was 56.3 ± 4.0 ms [161, 162] (Table 1, Fig. 5). One study reported T2 values in controls of 48.1 ± 3.2 ms suggesting significantly higher T2 values in HCM patients [161] (SMD = 1.95, 95% Cl [0.93, 2.97], I2 = N/A, P <  0.01, Fig. 16). In that same study [161] the T2 values were measured in the patient myocardium with visually high T2, which was present in 38% of the patients. For the patients without LGE in that study the myocardial T2 value of 48.8 ± 2.4 ms was not significantly different from controls. Furthermore, there were no studies performed with T2 values acquired at 3 T and insufficient studies were available for further analysis regarding covariates and publication bias.

Fig. 16
figure 16

Standardized mean differences between T2 of hypertrophic cardiomyopathy (HCM) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Dilated cardiomyopathy

The weighted mean T2* value at 3 T in DCM patients was 22.7 ± 3.6 ms [160, 163] and only one of those studies reported T2* values in controls of 21.0 ± 6.4 ms [160] (Table 1, Fig. 3). The random effects model was therefore only based on that study, and since that study reported T2* values of 18.7 ± 3.1 ms in DCM patients there was no significant change in T2* values (SMD = − 0.54, 95% Cl [− 1.09, 0.01], I2 = N/A, P = 0.06, Fig. 17). In both studies, patients had chronic established DCM and without myocarditis or other cardiomyopathies [160, 163]. Furthermore, there were no studies performed with T2* values acquired at 1.5 T and there were also insufficient studies available for further analysis regarding covariates and publication bias.

Fig. 17
figure 17

Standardized mean differences between T2* of dilated cardiomyopathy (DCM) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

The weighted mean T2 values at 1.5 T in DCM patients was 62.9 ± 5.7 ms and 55.4 ± 3.5 ms in controls [164,165,166,167,168,169] (Table 1, Fig. 5). At 3 T, these were 47.0 ± 5.0 ms in DCM patients and 45.0 ± 3.0 ms in controls [170] (Table 1, Fig. 6). All studies reported significantly higher T2 values in DCM patients compared to controls, except for the single study performed at 3 T [170]. The similar T2 values of patients and controls in this study might be related to their ROI placement, since they explicitly excluded positive LGE segments from the ROI, while all other studies used the entire myocardium without excluding positive LGE segments [164,165,166,167,168,169]. Nevertheless, the T2 values of positive and negative LGE segments were similar in all studies that reported T2 values of both segments [166,167,168]. The overall meta-analysis confirmed the significantly higher T2 values in DCM patients (SMD = 1.90, 95% CI [1.07, 2.72], P <  0.01, I2 = 89%, Fig. 18) and an exploratory meta-regression analysis indicated the MR vendor and the age difference between DCM patients and controls as possible covariates. The use of a Philips Healthcare CMR scanner and a bigger age difference between control and patient groups resulted in a larger SMD between DCM patients and controls.

Fig. 18
figure 18

Standardized mean differences between T2 of dilated cardiomyopathy (DCM) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Myocarditis

The weighted mean T2 values at 1.5 T in myocarditis patients was 61.9 ± 11.5 ms and 54.4 ± 5.9 ms in controls [25, 38, 171,172,173,174,175,176,177,178,179,180,181,182,183,184,185] (Table 1, Fig. 5). At 3 T, these were 63.8 ± 8.0 ms in myocarditis patients and 53.3 ± 3.3 ms in controls [186, 187] (Table 1, Fig. 6). The meta-analysis confirmed significantly higher T2 values in myocarditis patients (SMD = 1.33, 95% CI [1.00, 1.67], P <  0.01, I2 = 84%, Fig. 19). Multiple significant covariates were identified including; the difference in LVEF between patients and controls, the difference in percentage men between patients and controls, the time between symptoms and CMR, the number of echoes used in the CMR acquisition sequence, the CMR vendor and the slice thickness. These covariates together corrected for the total heterogeneity (I2 = 0%) and resulted in a larger SMD between myocarditis patients and controls when the same percentages of men was used in both groups, a significantly decreased LVEF was seen in patients, six echoes were acquired for the mapping, a Siemens Healthineers CMR vendor was used, a bigger slice thickness was used, and when the patients were scanned in the acute phase of myocarditis. Significant asymmetry was not found for either the random effects model (P = 0.12) or the mixed effects model (P = 0.10).

Fig. 19
figure 19

Standardized mean differences between T2 of myocarditis (MC) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

The time between symptom onset and CMR was found as significant covariate and therefore the population was divided between T2 values from patients in the acute phase and non-acute phase [192]. Acute myocarditis in patients was diagnosed using the European Society of Cardiology guideline [193] and these patients were referred for CMR shortly after symptom onset in the acute phase (< 14 days). Myocarditis patients in the non-acute phase either had chronic symptom duration (> 14 days) or underwent CMR follow-up. The weighted T2 value of myocarditis patients in the acute phase at 1.5 T was 63.5 ± 15.0 ms and at 3 T this was 63.8 ± 8.0 ms [25, 38, 167, 172,173,174,175,176,177,178,179, 181, 183,184,185,186,187] (Fig. 2, Supplementary Data 2). The weighted T2 value of myocarditis patients in the non-acute phase at 1.5 T was 58.3 ± 4.3 ms [173, 174, 179, 185] and at 3 T no T2 values were reported (Fig. 2, Supplementary Data 2). Furthermore, there were no studies that described T2* values for myocarditis.

Hypertension

One study reported T2* values at 1.5 T in hypertension patients of 26.3 ± 3.7 ms and 30.8 ± 2.7 ms in controls [188] (Table 1, Fig. 2). This suggested lower T2* values in hypertension patients, nevertheless this was not significant (SMD = − 1.46, 95% CI [− 3.21, 0.29], P = 0.10, I2 = 92%, Fig. 20). This study classified the included hypertension population in either presence of LVH or no presence of LVH, and showed in both subgroups lower T2* values, however in hypertension patients with LVH the T2* values were lowest. Furthermore, insufficient studies were available for further analysis regarding covariates and publication bias, and there were no studies that described T2* values acquired at 3 T or T2 results. Also, no published data was found on T2 or T2* for the cardiovascular risk populations obesity and diabetes.

Fig. 20
figure 20

Standardized mean differences between T2* of hypertension (HTN) patients and healthy controls with associated random effects weight factors. CI confidence interval, IV inverse variance

Full size image

Discussion

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, 164,165,166,167,168,169], 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 [155,156,157]. 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, 54,55,56, 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, 171,172,173, 175,176,177,178,179,180, 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, 60,61,62,63, 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, 127,128,129,130,131,132], 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, 71,72,73, 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.

Conclusion

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.