The study population consisted of adult patients (age greater than or equal to 18 years) who underwent heart transplantation at Mayo Clinic, Rochester, Minnesota, from January 1, 1997, through December 31, 2011. Patients were retrospectively identified by a search of our institutional patient database. All heart transplant patients during this time period who experienced severe biopsy-confirmed cell-mediated rejection (2R or greater) based on the 2004 International Society for Heart and Lung Transplantation (ISHLT) grading system [13] and who had LVEF greater than 55% were included in this study. Patients with known transplant vasculopathy were excluded. Transthoracic echocardiography was performed within 1 month prior to the biopsy. Control participants included heart transplant patients without transplant vasculopathy by angiography or intravascular ultrasonography, with no evidence of rejection by routine surveillance endomyocardial biopsy and normal echocardiography findings at one-year post-transplantation follow-up. Patients with abnormal renal function, significant valvular disease, or uncontrolled hypertension were excluded from the study. The electronic medical records were reviewed for demographic characteristics, along with clinical and echocardiographic data. Pertinent laboratory values, medications, and other clinical history were obtained at the time of echocardiography. All participants provided written informed consent. The study protocol was approved by the Mayo Clinic Institutional Review Board.

Comprehensive 2D echocardiography was performed using a Vivid E9 Ultrasound System (GE Healthcare) with a 2.5- to 4.0-MHz transducer, as part of the recommendations from the American Society of Echocardiography [14]. The modified Quinones equation or volumetric biplane Simpson method, or both, as appropriate, was used to quantify LVEF. Offline 2D-STE analysis (Velocity Vector Imaging) was performed retrospectively. Clinical, survival, and echocardiographic data were masked to the reviewers.

Three consecutive heart cycles from each of the apical views (apical 4-chamber, 3-chamber, and 2-chamber) were obtained for longitudinal strain analysis, including global longitudinal strain (GLS), global longitudinal strain rate (GLSR), and early diastolic global longitudinal strain rate (GLSRe). Short-axis views at the midpapillary level were obtained for circumferential strain analysis, including global circumferential strain (GCS) and global circumferential strain rate (GCSR).

Strain measurements were performed offline and analyzed with syngo Velocity Vector Imaging software (Siemens Medical Solutions USA, Inc) after transthoracic echocardiography evaluation. The LV wall was divided into 17 segments, and the tracing was performed using a 1-beat cycle starting at mid or end systole. The region of interest was adjusted to cover at least 90% of the myocardial wall thickness. Visual analysis was done first for endomyocardial tracking and, if required, manually adjusted. The epicardial border was then traced to include the entire myocardial thickness and was readjusted as needed. No segments were excluded for 2D-STE analyses. Clinical, survival, and echocardiographic data were masked to the reviewers.

Continuous variables are presented as mean (SD); categorical variables are presented as number (percentage of total). Continuous data were analyzed with a t test or a Wilcoxon rank sum test when assumptions were not met. Paired t tests were used to compare means for GLS and segmental longitudinal strain between groups. Categorical variables were analyzed using either the Pearson χ² or Fisher exact test. Receiver operating characteristic analysis was used to evaluate accuracy of strain and systolic strain rates to detect rejection.

For interobserver variability, two independent investigators (A.S.T. and S.B.-G.) to whom the initial results were masked assessed strain in 6 randomly selected patients. For intraobserver variability, repeat assessment of 2D-STE examination in 6 randomly selected patients by the same investigator was performed. To assess agreement, the concordance correlation coefficient (CCC) and Bland-Altman analysis, and calculation of bias (mean [SD] difference between 2 measurements) and of differences in individual pairs of measurements (limits of agreement) were used. Spearman rank correlation coefficient was also calculated.

A total of 65 heart transplant patients were included, 25 with severe rejection and 40 controls without rejection. Baseline demographics, risk factors, and clinical variables are summarized in Table 1. The mean (SD) age was 48.5 (13.5) years in the rejection group and 53.2 (11.5) years in the control group. There were more men than women in both groups (64 and 75%, respectively). All baseline variables, including body mass index, body surface area, hypertension, low-density lipoprotein cholesterol, diabetes mellitus, hypertension, heart rate, and hyperlipidemia, were similar in both groups. There were no significant differences in medication usage, including immunosuppression, prior prednisone use in the preceding 12 months, angiotensin-converting enzyme inhibitors, beta-blockers and statins. Eighty percent of patients had non-ischemic dilated cardiomyopathy as their heart transplant indication. In regards to symptoms, of the patients with severe rejection, 19 patients (76%) were asymptomatic at the time of diagnosis of severe rejection. The most frequently reported symptoms included lower extremity edema, dyspnea, dizziness, fatigue and headaches.

Both groups had preserved and similar mean (SD) LVEF (64.3% [5.2%] rejection group vs 64.5% [3.7%] control; P = .87). LV end-diastolic diameter (45.9 [4.6] mm vs 46.3 [3.7] mm; P = .75), ratio of early and late diastolic waves of mitral inflow, or E/A ratio (2.21 [1.03] vs 1.97 [0.70]; P = .51), and deceleration time of mitral inflow, or DT (180.3 [25.5] ms vs 169.0 [27.7] ms; P = .23) were not significantly different between the rejection and control group, respectively. Table 2 summarizes the echocardiographic data for both groups.

There were no significant differences in mean (SD) GLS (− 13.18% [3.32%] vs − 13.63% (2.35%); P = .52) and GLSR (− 0.86 [0.24] s^− 1 vs − 0.82 [0.15] s^− 1; P = .46) between the rejection and control groups. Strain analysis showed significant differences in mean (SD) GLSRe (1.04 [0.28] s^− 1 vs 0.80 [0.21] s^− 1; P = .02), GCS (− 15.74% [3.40%] vs − 19.90% [4.25%]; P < .001), and GCSR (− 1.09 [0.30] s^− 1 vs − 1.28 [0.30] s^− 1; P = .02) between the rejection and control groups, respectively. The area under the receiver operating characteristic curve for GCS was 0.77. With values less negative than a cutoff of − 17.60%, the sensitivity and specificity of GCS to detect severe acute rejection were 81.8 and 68.4%, respectively (Fig. 1).

Interobserver data between 2 independent researchers showed strong agreement for GLS and GCS. For GLS, the Bland-Altman plots showed statistically significant minimal bias (mean [SD], − 0.17 [3.6]; P < .05) with relatively narrower limits of agreement (− 7.3 to 6.9 [1.96SD]); CCC, 0.57 (95% CI, 0.19–0.80); Pearson ρ (precision), 0.57; and bias correction factor Cb (accuracy), 0.99. For GCS, the Bland-Altman plots showed statistically significant minimal bias (mean [SD], 0.14 [4.3]; P < .05) with relatively narrower limits of agreement (− 4.4 to 4.6 [1.96SD]); CCC, 0.66 (95% CI, − 0.11-0.93); Pearson ρ (precision), 0.69; and bias correction factor Cb (accuracy), 0.95. Intraobserver variability analysis for GCS and GLS also showed agreement: CCC of 0.85 (95% CI, 0.65–0.94) and 0.97 (95% CI, 0.81–0.99), respectively.