Left ventricular subclinical myocardial dysfunction in uncomplicated type 2 diabetes mellitus is associated with impaired myocardial perfusion: a contrast-enhanced cardiovascular magnetic resonance study

The study complied with the mandate of the Declaration of Helsinki and was approved by the Institutional Review Board of our hospital with written informed consent obtained from all study participants. From April 2016 to November 2017, we prospectively recruited 78 patients with T2DM from outpatients attending the Department of Endocrinology at our institution. T2DM was diagnosed according to the current American Diabetes Association guidelines [17]. The main exclusion criteria were clinical evidence of CAD, cardiomyopathy, myocardial infarction, or valvular heart disease (confirmed by echocardiography, electrocardiogram, or coronary computed tomographic angiography), a history of chest pain, uncontrolled hypertension (systolic blood pressure > 160 mmHg), severe renal failure (estimated glomerular filtration rate, eGFR < 30 ml/min), or contraindications to MR imaging. For comparison, 31 age- and sex-matched healthy volunteers with no history of cardiac disease or diabetes mellitus were recruited from the local community. All apparently healthy individuals who were to be recruited as controls underwent laboratory measurements before enrollment, and the exclusion criteria were impaired fasting glucose (fasting glucose > 6.1 mmol/l), dyslipidemia, or hypertension (blood pressure > 140/90 mmHg).

The height and weight of all participants were recorded and BMI was calculated as weight (kg) divided by the square of height (m). The duration of diabetes was recorded as reported by the patient. Blood pressure was recorded as an average of three measurements in the right arm in a sitting position that were obtained after a 10 min resting period. Before CMR imaging, fasting (8 h) blood samples were obtained from patients and controls for standard laboratory analysis, including plasma glucose, glycosylated hemoglobin, and blood lipids, according to standard procedures of the central clinical laboratory in our hospital.

CMR imaging was performed with patients in the supine position and on a 3.0-T whole-body scanner (Trio Tim; Siemens Medical Solutions, Erlangen, Germany) with a dedicated two-element cardiac-phased array coil. A standard ECG-triggering device was also simultaneously used, and data was acquired during the breath-holding period. After the cardiac axes were determined using localizers, a balanced steady-state free-precession (bSSFP) sequence (TR/TE 39.34/1.22 ms, flip angle 38°, slice thickness 8 mm, field of view 360 × 300 mm, matrix size 256 × 166) was used to obtain cine images of the long-axis and the short-axis views and to achieve complete LV coverage from the mitral valve to the apex. For perfusion imaging, a contrast dose of 0.2 ml/kg gadobenate dimeglumine (MultiHance; Bracco, Milan, Italy) was administered using an automated injector (Stellant, MEDRAD, Indianola, PA, USA) at a flow rate of 2.5–3.0 ml/s, followed by a 20 ml saline flush at a rate of 3.0 ml/s. Rest perfusion images were acquired concurrently with intravenous contrast agents in three standard short-axis slices (apical, middle, and basal) and in one 4-chamber view slice using an inversion-recovery echo-planar imaging sequence (TR/TE 163.00/0.98 ms, flip angle 10°, slice thickness 8 mm, field of view 360 × 270 mm, matrix size 256 × 192). To exclude myocardial infarction, late gadolinium enhancement (LGE) imaging (Additional file 1) was obtained at an average of 10–15 min after contrast injection by using a segmented-turbo-FLASH–phase-sensitive inversion recovery (PSIR) sequence (TR/TE = 750.00/1.18 ms; flip angle = 40°; slice thickness = 8 mm; field of view = 400 × 270 mm²; matrix size = 184 × 256).

All MRI data were uploaded to commercially software (cmr42 version 5.9.1, Circle Cardiovascular Imaging Inc., Calgary, Canada), the measurements were analyzed by a investigator who was blinded to the status (DM vs control, newly diagnosed vs. longer-term DM) of the subjects. Left ventricular function parameters, namely, LV end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), and ejection fraction (EF) were calculated by manually tracing the endocardial and epicardial contours in serial short-axis slices at the end-diastolic and end-systolic phases. For analysis of LV myocardial strain, long-axis 2-chamber, 4-chamber, and short-axis slices were loaded into the 3-dimensional (3D) tissue tracking module [18]. The endocardial and epicardial contours were delineated manually per slice at end-diastole phase (reference phase) in all series, while carefully excluding the papillary muscles and moderator bands. LV global myocardial strain parameters were acquired automatically, including radial, circumferential, and longitudinal peak strain (PS), peak systolic strain rate (PSSR), and peak diastolic strain rate (PDSR). For analyzing LV myocardial perfusion, endocardial and epicardial contours were delineated manually in the first-pass perfusion images of the basal, mid, and apical short-axis slices, along with a region of interest drawn in the LV blood pool. Conforming to AHA standard segmentation recommendations, a 16-segment mode (Bull’s eye plot) was constructed, which included six basal segments, six middle segments, and four apical segments [19]. Subsequently, a myocardial signal intensity-time curve was generated (Fig. 1), and LV segmental perfusion parameters such as upslope, time to maximum signal intensity (TTM), and max signal intensity (Max SI) were obtained automatically. For each subject, all LV global perfusion parameters were calculated using average regional values for the 16 myocardial segments.

To determine intraobserver variability, LV global deformation and perfusion parameters in 30 random cases that included 22 T2DM patients and 8 normal controls were measured twice in 2-week intervals by a radiologist. Then, a second investigator, who was blinded to the first investigator’s results, reanalyzed the measurements. Finally, the interobserver variability was assessed on the basis of the two investigators’ results.

All statistical analyses were performed with SPSS statistics for Windows (Version 17.0; SPSS Institute, Inc., Chicago, IL, USA). All data were evaluated for normality using the Kolmogorov–Smirnov test and are presented as mean ± standard deviation. Homogeneity of variance was evaluated using the Levene’s test. One-way analysis of variance (one-way ANOVA) was used to compare LV function, deformation, and perfusion parameters among controls, newly diagnosed diabetics, and longer-term diabetics, while, Least-Significant-Difference was used to analyze the difference between pooled data of the diabetes and normal groups, and the Student–Newman–Keuls or Least—Significant Difference test was used to evaluate the difference within the diabetes groups when the p-value of one-way ANOVA was less than 0.05. The Kruskal–Wallis rank test was used to analyze parameters that did not conform to normality or show homogeneity of variance. Binary variables were analyzed using the cross tabs Chi square test. Pearson’s correlation was used to examine the correlation between LV strain and perfusion parameters. The intraclass correlation coefficient (ICC) was used to evaluate both inter- and intra-observer variability.

We recruited 109 subjects (78 patients and 31 controls) for the study and acquired CMR imaging data for all subjects. However, data from five subjects had to be excluded because of poor image quality (e.g., severe motion artifacts in four T2DM patients and one control) and from three patients because the inferior quality of the perfusion data sets prevented quantitative analysis. Thus, the final study cohort comprised 71 T2DM patients and 30 healthy controls (proportion of males 56.7% vs. 57.7%, p = 0.92; mean age 53.99 ± 11.15 vs. 53.23 ± 8.59 years, p = 0.74). Based on the duration of diabetes [7, 20], all T2DM patients were categorized as either newly diagnosed DM (duration of diabetes ≤ 5 years, n = 31) or longer-term DM (duration of diabetes > 5 years, n = 41). The baseline characteristics of the T2DM patients and healthy volunteers are presented in Table 1. Compared to the newly diagnosed group, characteristics such as proportion of males, BMI, and systolic blood pressure were statistically higher in the longer-term DM cohort, while systolic blood pressure was significantly higher in the newly diagnosed DM group compared to the control group (p < 0.05). Fasting plasma glucose, glycated hemoglobin, and triglyceride levels were higher in patients with T2DM compared to healthy subjects but were not different between the two subgroups of T2DM patients. As expected, diabetic patients had lower high-density lipoprotein (HDL) and lower low-density lipoprotein compared to healthy controls. Total cholesterol was also significantly lower in newly diagnosed DM patients than in control subjects, which might be due to the fact that a significant proportion of diabetics were also on statin therapy.

CMR imaging results for LV volume and function are summarized in Table 2. There were no significant differences in LVEDVI (LVEDV index), LVESVI (LVESV index) and LVEF among control subjects, newly diagnosed diabetics, and longer-term diabetics. However, global longitudinal PS was significantly lower in longer-term DM compared to normal subjects (p < 0.017). Within the T2DM patient subgroups, the global longitudinal PSSR was significantly lower in the longer-term DM group than the newly diagnosed group (p < 0.017). Additionally, T2DM patients showed lower global radial, circumferential, and longitudinal PDSR compared to control subjects (all p < 0.017), except for radial PDSR in the newly diagnosed diabetics (p > 0.017). Within the two subgroups of T2DM patients, the global radial, circumferential, and longitudinal PDSR were also significantly lower in the longer-term group compared to the newly diagnosed group (p < 0.017 for all; Fig. 2).

Data on global first-pass perfusion parameters for all subjects are presented in Table 2. Compared to the normal subjects, the upslope in the T2DM patients was significantly reduced (p < 0.017). Next, the TTM was higher in the longer-term DM group than the normal subjects (37.85 ± 11.03 vs. 31.45 ± 5.45, p < 0.017). Additionally, Max SI was significantly reduced in the longer-term diabetics compared to the newly diagnosed diabetics and the normal subjects (p < 0.017 for all; Fig. 3).

Multivariable linear regression analysis revealed that considering covariates (systolic blood pressure, age, sex, and BMI), T2DM was independently associated with longitudinal PS (β = 0.263, p = 0.007, model R² = 0.11), the PSSR (β = 0.243, p = 0.014, model R² = 0.05), and all three directions (radial, circumferential, and longitudinal) of the PSSR (β = 0.255, p = 0.008, model R² = 0.05; β = 0.0.560, p < 0.001, model R² = 0.31; and β = 0.657, p < 0.001, model R² = 0.43, respectively). T2DM was also found to be independently associated with first-pass perfusion parameters (upslope: β = − 0.399, p < 0.001, model R² = 0.15; Max SI: β = –0.316, p < 0.001, model R² = 0.09), except for TTM (β = 0.137, p = 0.195, model R² = 0.11).

As shown in Table 3, there was a weak correlation between decreased LVEF and increased TTM in the T2DM patients (r = − 0.306, p = 0.01). LV global radial PS and PSSR, and circumferential PDSR were inversely correlated to TTM (r = − 0.334, p = 0.004; r = − 0.342, p = 0.004; r = − 0.367, p = 0.002, respectively) in patients with T2DM. In contrast, longitudinal PS, circumferential PSSR, and radial PDSR were positively associated with TTM (r = 0.378, p = 0.001; r = 0.355, p = 0.002; r = 0.389, p = 0.001, respectively). Further, longitudinal PSSR was significantly associated with upslope (r = − 0.346, p = 0.003) and TTM (r = 0.515, p < 0.001; Fig. 4).

Multivariable analysis linear regression demonstrated that TTM was independently associated with longitudinal PS (β = 0.378, p = 0.001, model R² = 0.13) and PSSR (β = 0.466, p < 0.001, model R² = 0.33). In addition, the upslope was independently associated with the longitudinal PSSR (β = − 0.291; p = 0.009, model R² = 0.19; Table 4).

Table 5 summarizes the inter- and intra-observer variability for feature tracking and first-pass perfusion analysis. The ICCs for intra- and interobserver variability were 0.828–0.952 and 0.777–0.925, respectively, in feature tracking and 0.979–0.987 and 0.977–0.983, respectively, in first-pass perfusion, suggesting that both techniques are in agreement.