This was a single-center retrospective study cohort. Sixty-three consecutive patients with a clinical diagnosis of TA and who underwent CMR examination from January 2015 to October 2022 were retrospectively enrolled in a single-center institution (Beijing Anzhen Hospital affiliated with Capital Medical University). Diagnosis of TA relied on the diagnostic criteria established by the American College of Rheumatology in 1990 (18). The study was performed in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Beijing Anzhen Hospital affiliated with Capital Medical University. Written informed consent was obtained from all patients with TA.

Demographic characteristics, hemodynamic indices at admission and clinical history were collected in the electronic medical record system for all TA patients.

CMR examination was performed on 3 Tesla scanners (MAGNETOM Verio, Siemens Healthcare, Erlangen, Germany; MR750W, General Electric Healthcare, Waukesha, WI, USA; Ingenia CX, Philips, Best, The Netherlands) in a single-center institution (Beijing Anzhen Hospital affiliated with Capital Medical University). All patients were trained in breathing before the CMR examination. ECG gating and dedicated coil are routinely used during CMR examination. CMR scanning protocols comprised conventional cine sequences and LGE sequences. Left ventricular (LV) short-axis from the mitral orifice to the apical range and long-axis (four-chamber, three-chamber, and two-chamber views) cine images were acquired during breath-holding. The parameters for short- and long-axis cine sequences are as follows: field of view (FOV) = 340 × 314 mm, matrix = 256 × 256, repetition time (TR)/echo time (TE) = 4.1/1.35 ms, slice thickness = 8 mm, gap = 0 mm (MAGNETOM Verio); FOV = 300 × 300 mm, matrix = 256 × 256, TE/TR = 3.8/0.5 ms, slice thickness = 8 mm, gap = 0 mm (MR750W); FOV = 270 × 270 mm, matrix = 152 × 119, TE/TR = 2.7/1.37 ms, slice thickness = 8 mm, gap = 0 mm (Ingenia CX).

LGE images covering the left ventricle in short-axis and long-axis views were obtained at 10–15 min after intravenous injection of gadopentetate dimeglumine (Bayer Healthcare) at a dose of 0.1 mmol per kilogram of body weight. The inversion time count sequence was used to find the optimal inversion time (TI) before acquiring the LGE images. The optimal TI was defined as the TI of myocardial signal nulling. Based on the adjusted optimal TI time, LGE images were acquired using a phase-sensitive inversion recovery (PSIR) sequence. The parameters of the PSIR sequence are as below: FOV = 340 × 285 mm, matrix = 256 × 256, TE/TR = 15.60/1.56 ms, slice thickness = 8 mm, gap = 0 mm (MAGNETOM Verio); FOV = 300 × 300 mm, matrix = 256 × 256, TE/TR = 6.20/0.5 ms, slice thickness = 8 mm, gap = 0 mm (MR750W); FOV = 360 × 270 mm, matrix = 256 × 192, TE/TR = 12.44/1.19 ms, slice thickness = 8 mm, gap = 0 mm (Ingenia CX).

LV wall thickness (LVWT) was measured online at the left ventricular end-diastolic mid-segment septal wall on a short-axis view through the Radiology work system. The CMR short-axis cine images were imported offline into the commercial software (CVI42, version 5.14, Circle Cardiovascular Imaging Inc., Calgary, Canada). The end-systolic and end-diastolic phases of both ventricles were detected and identified semi-automatically. The LV endocardial and epicardial outlines were sketched to calculate volume, LV ejection fraction (LVEF), and myocardial mass on a short-axis cine view. Myocardial trabeculae and papillary muscles were counted in volume but excluded from the myocardial mass. Only the right ventricular (RV) endocardial contour was drawn to measure RV volume and EF. However, all volumetric and myocardial mass parameters needed to be normalized by dividing by body surface area.

All LGE images were visually evaluated online by two independent double-blind cardiovascular diagnostic radiologists (D.T.L. and J.Y.L.) with more than ten years of experience in the Radiology work system. LGE-positive was defined as hyperintensity within the myocardium was visually observed in any of the sixteen segments of the left ventricle. According to the findings of visual observation, all patients with TA were divided into the LGE-negative and LGE-positive groups. The location, distribution, and pattern of LGE in the sixteen segments of the left ventricle of patients with LGE-positive were visually assessed and recorded. In case of inconsistency, a third cardiovascular diagnostic radiologist (Z.Y.W.) with more than fifteen years of experience visually observed and made a final decision. LGE-positive was categorized into four types according to visually observed distribution of LGE-positive: subendocardial LGE, middle wall LGE, subepicardial LGE, and transmural LGE. LGE-positive was classified into three types based on visually observed patterns of LGE-positive: line LGE, patchy LGE, and infarcted LGE pattern. Representative examples of LGE-positive images are displayed in Figure 1. In addition, the number of LGE-positive segment based on visual observation in each TA patient was also recorded.

The primary outcome of this study was the occurrence of major adverse cardiovascular events, defined as a composite of cardiovascular death, hospitalization for heart failure, coronary artery revascularization, and stroke. The occurrence of any of the events listed above is defined as reaching the primary outcome. If patients had multiple adverse cardiovascular events, only the first was selected as the primary outcome. The cardiovascular events were obtained by telephone contact or reviewing medical records for overall TA patients. Furthermore, the follow-up time from the date of the CMR examination to the presence of adverse cardiovascular events was calculated.

All continuous data were tested for normality distribution using the Shapiro-Wilk test. Continuous data obeying a normal distribution are expressed as mean ± SD, otherwise as median [interquartile range (IQR)]. Categorical variables are presented as numbers and percentages. The difference between the continuous data of the LGE-negative and LGE-positive groups was compared using the unpaired t-test or the Mann–Whitney U test (two sides). The difference between the categorical variables of the two groups was analyzed using the Chi-square test or Fisher exact test (two sides). Univariate and multivariate Cox proportional hazards regression analyses were used to assess the effect of various variables on the primary outcome. The association of each variable with outcomes was first evaluated in a univariate analysis by applying a forward stepwise regression approach. Variates with P < 0.05 were selected in the univariate analysis and were included in the further multivariate analysis. In order to avoid collinearity between CMR functional parameters (including LVEF, LVEDVI, and LVESVI), the variance inflation factor (VIF) was calculated. Only factor with VIF less than 5 was considered for included in the further multivariate analysis. Hazard ratios (HRs) along with their 95% confidence intervals (CIs) were also calculated. Receiver operating characteristic (ROC) curves were adopted to determine the diagnostic efficacy of CMR parameters to predict primary outcomes. The optimal cut-off values were sought by calculating the Youden index. All participants were divided into higher and lower risk groups according to the optimal cut-off values. A comparison between the higher and lower risk groups was assessed using the log-rank test. The event-free survival probability was plotted using the Kaplan–Meier curve. All statistical analyses were performed on SPSS software (version 26.0, International Business Machines, Armonk, New York, USA). For all calculations, a statistically significant difference was defined as P < 0.05 (two sides).

The study flowchart of patient enrollment is presented in Figure 2. A total of 63 consecutive patients with TA undergoing a CMR examination between January 2015 and October 2022 were retrospectively enrolled. Two patients with non-contrast CMR due to renal dysfunction and one patient lost to follow-up were excluded. Eventually, sixty patients with TA were included in the study cohort.

LGE-positive was found in twenty-one patients (21/60, 35%) with TA, while the remaining were LGE-negative (39/60, 65%). Clinical baseline characteristics of the overall study group, LGE-negative group, and LGE-positive group were summarized in Table 1. The age of total TA patients was 38.2 ± 13.8 years and was predominantly female (54/60, 90.0%). Patients in the LGE-positive group were older compared with patients in the LGE-negative group (44.5 ± 16.3 vs. 34.9 ± 11.1 years; P = 0.027). No statistically significant differences were detected between the LGE-negative and LGE-positive groups regarding demographics and clinical parameters (all P > 0.05).

Morphological and functional parameters of CMR in the overall study group, LGE-negative group, and LGE-positive group are displayed in Table 2. No statistically significant difference (P = 0.089) was detected in LVWT between the LGE-negative and LGE-positive groups. For LV functional parameters, lower LVEF (P = 0.033), higher LV end-diastolic volume index (LVEDVI; P = 0.033) and end-systolic volume index (LVESVI; P = 0.012) were observed in the LGE-positive group compared with the LGE-negative group. Moreover, increased LV mass index (LVMI; P = 0.008) was also noticed in the LGE-positive group compared with the LGE-negative group. However, LV stroke volume index and RV functional parameters did not exhibit statistically significant differences (all P > 0.05) between the LGE-negative and LGE-positive groups.

The general characteristics of LGE in the overall study group and LGE-positive group are shown in Table 2. There were 6.0 (IQR: 2.0–9.0) LGE segments per TA patient in the LGE-positive group. For the location of the LGE, the LGE was observed in the anterior wall (9/21, 42.9%), the septal wall (18/21, 85.7%), the inferior wall (9/21, 42.9%), and the lateral wall (12/21, 57.1%), with the septal wall predominantly. For the distribution of LGE, middle wall LGE was observed in twelve of twenty-one patients (12/21, 57.1%), followed by subendocardial LGE (7/21, 33.3%), transmural LGE (5/21, 23.8%), and subepicardial LGE (1/21, 4.8%). For the pattern of LGE, line, patchy, and infarction patterns were observed in five (5/21, 23.8%), seven (7/21, 33.3%), and ten (10/21, 47.6%) of twenty-one patients, respectively.

The sixteen segment characteristics of LGE in the LGE-positive group are displayed in Figure 3. For the patients of the LGE-positive group, LGE-positive was detected in 131 (131/336, 39.0%) of the total 336 segments. In terms of distribution, subendocardial LGE and middle-wall LGE are more common. Moreover, the prevalence of patchy LGE and infarction LGE patterns was higher compared with line pattern in the LGE-positive group.

During a median follow-up period of 1,892 days (IQR: 1,764–1,988 days), the primary outcomes were recorded in thirteen patients with TA (13/60, 21.7%), with cardiovascular death occurred in four patients (4/60, 6.7%), hospitalization for heart failure in five patients (5/60, 8.3%), coronary revascularization in four patients due to coronary atherosclerotic disease (4/60, 6.7%), and no patient experienced stroke (0/60, 0%).

Univariate and multivariate Cox proportional hazards regression analyses for primary outcomes in the total patients with TA are displayed in Table 3. Univariate Cox regression analysis showed that the LVWT (HR = 1.180, 95%CI: 1.019–1.388; P = 0.028), LVEF (HR = 0.949, 95%CI: 0.920–0.979; P = 0.001), LVEDVI (HR = 1.014, 95%CI: 1.001–1.027; P = 0.040), LVESVI (HR = 1.016, 95%CI: 1.004–1.029; P = 0.003), LVMI (HR = 1.030, 95%CI: 1.013–1.048; P = 0.001), and LGE-positive (HR = 4.478, 95%CI: 1.376–14.570; P = 0.013) were independently associated with the primary outcomes, respectively.

The determination of cut-off values for the CMR parameters that were statistically significant in the univariates analysis (including only LVWT, LVEDVI, LVESVI, and LVMI) was displayed in Supplementary Table S1. After that, survival curves were plotted based on the cut-off values of LVWT, LVEDVI, LVESVI, and LVMI divided into two subgroups of higher and lower levels. Similarly, subgroups based on LVEF and the status of the LGE were also categorized into LVEF ≥ 50% vs. <50% and those with LGE-negative vs. LGE-positive, respectively. The Kaplan–Meier curves drawn for the above subgroups are shown in Figure 4.

After adjustment for age and gender, we included LVWT, LVEF, LVMI, and LGE-positive as covariates in the multivariate analysis. However, only LVMI (HR = 1.030, 95%CI: 1.013–1.048; P = 0.001) maintained an independent predictor of the primary outcome in further multivariate analysis. Based on the optimal predictive value of LVMI, all TA patients were categorized into two subgroups with LVMI < 57.5 g/m² vs. ≥57.5 g/m². The Kaplan–Meier survival plot based on subgroups of LVMI was also depicted in Figure 4. The event-free survival probability was significantly lower for TA patients with LVMI ≥ 57.5 g/m² compared with patients with LVMI < 57.5 g/m² (log-rank P < 0.001).

Up to now, numerous studies have emerged regarding the value of LGE in the prognostic assessment of cardiovascular disease. Although there are discrepancies between the results of different studies, most have shown that LGE is an independent predictor of adverse cardiovascular events. This has been confirmed in meta-analyses in the populations of suspected or known coronary artery disease (26), hypertrophic cardiomyopathy (27, 28), dilated cardiomyopathy (29–31), systemic amyloidosis (32), cardiac sarcoidosis (33, 34), aortic stenosis (35), and myocarditis (36). In this study, our findings showed that LGE was independently correlated with the primary outcomes in the univariate analysis. However, LGE did not maintain its value as an independent predictor of primary outcomes after various variates were added to the multivariate analysis.

In contrast, LVMI showed important value in providing independent prognostic information in patients with TA in the multivariate analysis. In terms of the predictive value of LGE, our findings are inconsistent with those of previous studies. Actually, this was a surprising and unexpected discovery for us. There are several possible reasons for this inconsistency. First, the sample size and adverse cardiovascular events of our study are relatively small. Therefore, LGE may not be statistically significant in multivariate analysis. Second, TA as a systemic and complex disease, cardiac abnormalities due to TA can manifest in different disease forms. The visual observation of LGE may be influenced by a variety of diseases. We speculate that LGE may reflect adverse and late myocardial remodeling, whereas LVMI may reflect compensatory and early myocardial remodeling. For the prognostic assessment of cardiovascular abnormalities due to TA, LVMI may be a more sensitive and stronger independent predictor of prognosis compared with LGE in patients with TA. However, this needs to be confirmed by further large-sample study cohorts and echocardiography studies.

We realize some limitations of this study. First, TA is a rare systemic disease and predominantly female, but this study only retrospectively included a relatively small number of patients with TA from one hospital, which may overestimate the occurrence of LGE-positive due to selection bias. Also, because of the limitations of sample size and number of outcomes, this study did not include too many factors to avoid the problem of overfitting in the univariate and multivariate Cox proportional hazards regression analyses. The predictive value of LGE needs to be further confirmed by prospective studies with multicenter and large samples. Second, although a blinded interpretation of LGE was applied, the difference in LGE image quality due to different scanner equipment and parameters may still exist. Third, LGE imaging has limitations in detecting diffuse myocardial fibrosis. Indeed, patients with LGE-negative may have mild diffuse myocardial fibrosis due to increased LV afterload. However, this requires further T1 mapping imaging and analysis. Unfortunately, most of the patients in this study did not undergo a T1 mapping scan. Therefore, T1 mapping remains promising for detecting early myocardial remodeling and prognostic assessment but awaits future investigation.