In normal subjects, left ventricular (LV) dimensions have been shown to decrease over time, while wall thickness is increasing. The aim of this study was to investigate LV remodeling in a cohort of patients with type 2 diabetes mellitus during a 3-year follow-up period and its potential association with decreased longitudinal systolic strain (ε L ).
One hundred seventy-two patients with type 2 diabetes without overt heart disease were prospectively enrolled and underwent echocardiography with speckle-tracking imaging to assess global LV ε L at baseline and at 3 years. The associations between alteration in ε L (defined as |ε L | < 18%), LV geometry at baseline, and LV remodeling over time were evaluated.
Among the 172 enrolled patients, 154 completed 3-year follow-up. At baseline, patients with ε L alteration had higher LV end-systolic volumes (28 ± 11 vs 23 ± 9 mL, P < .001) and relative wall thicknesses (RWT; 0.44 ± 0.06 vs 0.40 ± 0.07, P = .008) compared with those with normal ε L . At 3-year follow-up, RWTs remained stable in both groups. LV volumes significantly decreased in patients with normal ε L but not in patients with ε L alteration. Multivariate analysis showed that ε L alteration was independently associated with LV end-systolic volume (β = 5.0, P = .006) and RWT (β = 0.03, P = .03) at baseline and with changes in both LV end-diastolic volume (β = 19.1, P = .001) and LV end-systolic volume (β = 2.6, P = .047) over 3 years.
In patients with type 2 diabetes, ε L alteration was associated with higher RWT and LV volumes and with the absence of decreases in LV volumes over time, which might be an early sign of adverse LV remodeling.
Diabetes mellitus is associated with an increased risk for heart failure, even in the absence of coronary artery disease or hypertension, because diabetes itself is responsible for the development of diabetic cardiomyopathy. This pathology is responsible for increases in both cardiovascular morbidity and mortality.
We and others have shown that patients with diabetes exhibit decreased left ventricular (LV) systolic strain compared with euglycemic subjects. It was speculated that such an abnormality could be considered an early marker of diabetic cardiomyopathy. However, the potential impact of these subtle abnormalities on the evolution of LV function and LV geometry remains unknown. Indeed, the association between abnormal systolic strain in patients with diabetes and cardiac remodeling over time has never been investigated.
Recently, large cohort studies have underlined the influence of diabetes on cardiac LV remodeling over the lifetime. Although in normal subjects, the aging process is associated with a progressive increase in LV wall thickness and a decrease in LV cavity dimensions, the presence of diabetes induces a more pronounced increase in LV wall thickness but the absence of a proportional decrease in cavity dimensions.
We hypothesized that in patients with type 2 diabetes mellitus, alterations in longitudinal systolic deformation are associated with LV remodeling. Therefore, the aim of this study was to investigate LV remodeling in a cohort of patients with type 2 diabetes mellitus during a 3-year follow-up period and its potential association with alteration in longitudinal systolic strain (ε L ).
Between February 2006 and June 2009, 172 consecutive patients with type 2 diabetes referred to the outpatient Department of Endocrinology at our institution were prospectively included. The inclusion criteria were (1) age between 35 and 75 years, (2) oral antidiabetic or insulin treatment, and (3) LV ejection fraction (LVEF) > 50%. Exclusion criteria were (1) symptoms, signs (clinical or electrocardiographic), or history of heart disease; (2) presence of regional LV wall motion abnormalities; (3) absence of sinus rhythm; (4) history of cardiomyopathy, coronary artery disease, or valvular heart disease; (5) severe renal failure, defined as creatinine clearance < 30 mL/min; (6) echocardiographic images unsuitable for quantification; (7) severely uncontrolled diabetes, defined as glycosylated hemoglobin (HbA 1c ) > 12% or glycemia > 3 g/L; and (8) uncontrolled blood pressure at rest (defined as blood pressure > 180/100 mm Hg). All patients underwent exercise stress tests, stress echocardiography, or myocardial perfusion scintigraphy within the month before inclusion to exclude silent ischemia.
Among the 172 enrolled patients, seven declined to repeat the echocardiographic examination, two were censored because of cancer treated with chemotherapy, two died (one sudden death and one pancreatic cancer), one had stress cardiomyopathy, and six had nonfatal myocardial infarctions and/or underwent revascularization. The remaining study population consisted of 154 patients.
All subjects provided informed consent to participate, and the study was approved by the ethics committee of our institution.
All patients underwent physical examinations, standard echocardiography, and biochemical analysis on the same day at baseline and after 3 years.
Resting transthoracic echocardiography was performed in the left lateral decubitus position using a commercially available ultrasound system (Vivid 7 or 9; GE Medical Systems, Oslo, Norway). All acquisitions were digitally stored in raw-data format from at least three consecutive heartbeats for offline analysis (EchoPAC; GE Vingmed Ultrasound AS, Horten, Norway), which was performed by two experienced observers blinded to the other data (L.E., C.B.).
LV wall thickness was measured from M-mode images from the parasternal long-axis view according to the recommended criteria. Total LV wall thickness was calculated as the sum of septal wall thickness and posterior wall thickness. LV mass was determined as recommended using Devereux’s formula and indexed to body surface area. Relative wall thickness (RWT) was calculated as (2 × posterior wall thickness at end-diastole)/LV end-diastolic diameter.
LV end-diastolic volume (LVEDV) and LV end-systolic volume (LVESV) and LVEF were calculated from the apical four-chamber and two-chamber views using the modified biplane Simpson’s method.
Using pulsed-wave Doppler, mitral inflow velocities, peak early (E) and late (A) diastolic velocities, the E/A ratio, and E-wave deceleration time were measured. The annular early diastolic velocity (e′) was assessed at the lateral and septal sites of the mitral annulus using pulsed-wave Doppler tissue imaging. The average e′ value (from the lateral and septal sites) was used to calculate the E/e′ ratio. Left atrial area was measured in an apical four-chamber view using planimetry, and left atrial volume was assessed as previously described.
Strain Analysis by Speckle-Tracking Imaging
Speckle-tracking analysis was performed using a dedicated software package (EchoPAC). Peak ε L was measured in the apical four-chamber and two-chamber views (frame rate, 70–80 frames/sec) The endocardial border was manually traced from an end-systolic frame. The software automatically detected the epicardial border, and the region of interest was manually adjusted to include the entire myocardial wall. Thus, the software tracked the contour throughout the entire cardiac cycle frame by frame. The quality of tracking was verified both automatically and visually, and the region of interest was modified and corrected by the observer if judged necessary to obtain optimal tracking. The software automatically divided the LV walls into six segments for each view and calculated the segmental strain values. Each segmental peak ε L value was collected, and the average of longitudinal segmental strain values was calculated for each patient and presented as ε L .
Alteration in ε L was defined as |ε L | < 18% and normal ε L as |ε L | ≥ 18% (mean value − 2 standard deviations in normal subjects), as previously published by our group, in accordance with others and as reported in the American Society of Echocardiography and European Association of Echocardiography consensus statement on techniques for the quantitative evaluation of cardiac mechanics ( Figure 1 ).
To define reproducibility, 15 patients were randomly selected in the population study. In these patients, LV dimensional measurements by M-mode echocardiography, LV volumes, and ε L analysis were repeated 3 months apart by the same observer and performed by a second observer. The first observer (L.E.) was blinded to previous measurements during the second analysis, and the second observer (C.B.) was blinded to measurements of the first observer. A minimum of six cardiac cycles were available for each measurement (three cardiac cycles per loop and at least two loops for each view), and the reader was allowed to select the best cardiac cycle each time and to repeat and average the measurement if judged necessary.
Intraobserver and interobserver variability was calculated as the absolute difference divided by the average of the two measurements for each parameter and is expressed as a relative value (percentage of variability).
Blood samples were taken for the biochemical analysis of renal function, triglycerides, total cholesterol, and HbA 1c . Microalbuminuria was measured by immunonephelometry.
Analyses at Baseline
Normality of the continuous data was tested using a graphical method based on a histogram and quantile-quantile plot analysis. If non-normal distribution was suspected, we further performed a Kolmogorov-Smirnov test. Continuous and normally distributed data are presented as mean ± SD. Continuous data deviating from the normal distribution are presented as median (interquartile range), and categorical variables are presented as frequencies and percentages. Differences in baseline characteristics between patients with and without ε L alteration were tested using unpaired Student’s t tests for normally distributed data, Wilcoxon’s rank-sum tests for data deviating from the normal distribution, and χ 2 tests for categorical variables.
A simple linear regression analysis was used to assess associations between characteristics at baseline and LVESV and RWT. In the setting of linear regression, an increase of one unit of the independent variable (i.e., the explanatory variable) results in an increase of β of the dependent variable (i.e., the result variable) on an additive scale (i.e., β is added to the intercept). Association with the following baseline characteristics was tested for the two dependent variables (LVESV and RWT): age, gender, body mass index, diabetes duration, presence of a hypertension, dyslipidemia, smoking, peripheral artery disease, retinopathy, heart rate, insulin therapy, sulfonylureas, angiotensin-converting enzyme (ACE) inhibitor or angiotensin receptor blocker (ARB) therapy, HbA 1c , microalbuminuria, LV thickness, and ε L alteration. Then, backward selection was used to select variables independently associated with LVESV and RWT at baseline. To test whether the association between the presence of ε L alteration and LVESV varied according to gender, we introduced an interaction term, gender × ε L alteration, to the analysis. Variables associated with P values < .10 in a simple linear regression were selected to be included in the model.
Association between ε L Alteration and LV Remodeling during Follow-Up
Paired t tests for normally distributed data, Wilcoxon’s matched-pairs tests for data deviating from the normal distribution, and McNemar’s tests for categorical variables were used to assess changes between baseline and follow-up in each group according to the presence or the absence of ε L alteration.
The association between baseline characteristics and LV volume changes during follow-up was tested using bivariate linear regression with the absolute difference between LV volume at baseline and 3-year follow-up as the dependent variable. In these models, the tested baseline variable was entered as the independent variable along with LV volume at baseline. Association with the following prespecified baseline characteristics was tested: age, gender, body mass index, diabetes duration, presence of a hypertension, dyslipidemia, smoking, peripheral artery disease, retinopathy, insulin therapy, sulfonylureas, ACE inhibitor or ARB therapy, HbA 1c , microalbuminuria, LV thickness, LVEF, and ε L alteration. Then, backward selection was used to determine variables independently associated with changes in LV volume during the follow-up period. To test whether the association between ε L alteration and LV volumes varied according to gender, we introduced an interaction term, gender × ε L alteration, to the analysis. Variables associated with P values < .10 in the bivariate linear regression were selected to enter the model. The multivariate models were systematically adjusted for LV volume at baseline.
P values < .05 were considered statistically significant. Statistical analyses were performed using SPSS version 17.0.0 (SPSS, Inc, Chicago, IL).
Patients’ Characteristics at Baseline
The mean age of the study population was 58 ± 8 years, with a mean diabetes duration of 13 ± 8 years and a mean HbA 1c level of 7.7 ± 1.3%. Treatment included metformin for 110 patients (71%) and insulin therapy for 70 patients (45%) ( Table 1 ).
|Variable||Total population |
( n = 154)
|Normal strain (|ε L | ≥ 18%) |
( n = 118)
|Altered strain (|ε L | < 18%) |
( n = 36)
|Age (y)||58 ± 8||58 ± 8||57 ± 8||.51|
|Men||88 (57%)||60 (51%)||28 (78%)||.003|
|BMI (kg/m 2 )||29.5 ± 4.4||29.1 ± 4.3||30.6 ± 4.9||.09|
|Diabetes duration (y)||13 ± 8||13 ± 8||13 ± 7||.84|
|Treated hypertension||80 (52%)||58 (49%)||22 (61%)||.11|
|Dyslipidemia||90 (58%)||71 (60%)||19 (53%)||.30|
|Current smokers||26 (17%)||18 (15%)||8 (22%)||.21|
|Peripheral artery disease||51 (33%)||39 (33%)||12 (33%)||.37|
|Retinopathy||33 (21%)||25 (21%)||8 (22%)||.48|
|Systolic blood pressure (mm Hg)||132 ± 16||131 ± 16||135 ± 18||.20|
|Heart rate (beats/min)||75 ± 11||74 ± 12||78 ± 10||.07|
|Metformin||110 (71%)||87 (74%)||23 (64%)||.21|
|Sulfonylureas||64 (42%)||55 (47%)||9 (25%)||.02|
|Glitazones||33 (21%)||28 (24%)||5 (14%)||.17|
|Insulin||70 (45%)||50 (42%)||20 (56%)||.07|
|ACE inhibitors or ARBs||89 (58%)||61 (52%)||28 (78%)||.002|
|Statins||89 (58%)||69 (58%)||20 (56%)||.52|
|Antiplatelet agents||45 (29%)||32 (27%)||13 (36%)||.18|
|HbA 1c (%)||7.7 ± 1.3||7.8 ± 1.3||7.4 ± 1.4||.17|
|Triglycerides (mmol/L)||1.8 ± 1.4||1.8 ± 1.4||1.8 ± 1.4||.93|
|Total cholesterol (mmol/L)||4.7 ± 1.1||4.8 ± 1.1||4.5 ± 1.0||.17|
|eGFR (mL/min/1.73 m 2 )||83 ± 19||83 ± 22||83 ± 19||.96|
|Microalbuminuria (mg/L)||24 (11–69)||18 (10–60)||42 (15–148)||.045|
|Total LV wall thickness (mm)||20 ± 3||20 ± 3||21 ± 2||.12|
|LV mass index (g/m 2 )||93 ± 18||92 ± 19||95 ± 17||.48|
|RWT||0.41 ± 0.07||0.40 ± 0.07||0.44 ± 0.06||.008|
|LVEDV (mL)||79 ± 22||77 ± 20||85 ± 24||.08|
|LVEDVi (mL/m 2 )||42 ± 10||79 ± 22||79 ± 22||.50|
|LVESV (mL)||24 ± 9||23 ± 9||28 ± 11||<.001|
|LVESVi (mL/m 2 )||13 ± 5||12 ± 4||14 ± 5||.01|
|LVEF (%)||70 ± 7||71 ± 7||67 ± 7||.002|
|ε L (%)||−19.8 ± 2.4||−20.8 ± 1.8||−17.7 ± 1.2||—|
|LV diastolic function|
|E/A ratio||1.0 ± 0.2||1.0 ± 0.3||1.0 ± 0.2||.63|
|mDT (msec)||240 ± 54||240 ± 51||239 ± 61||.91|
|E/e′ ratio||9.7 ± 2.7||9.8 ± 2.8||9.3 ± 2.6||.32|
|LA volume (mL)||46 ± 13||45 ± 13||47 ± 14||.42|
Among the 154 patients, 36 (23%) had ε L alteration (defined as |ε L | < 18%). Diabetes duration was similar in the two groups. Blood pressure was in the normal range in both groups. No difference was observed regarding HbA 1c and renal function. Patients with ε L alteration were more likely treated with renin-angiotensin system inhibitors, were less likely to be treated with sulfonylureas associated with a trend toward more frequent use of insulin therapy, and had a higher level of microalbuminuria compared with patients with a normal ε L .
LV Geometry and Function at Baseline
Patients with ε L alteration had higher LVESV and LVESV indexed to body surface area than patients with normal ε L ( Table 1 ). In addition, despite similar LV wall thicknesses and LV mass indices in the two groups, RWTs were also higher in patients with ε L alteration compared with patients with normal ε L.
As determined by the inclusion criteria, LVEFs were in the normal range in both groups but were lower in patients with ε L alteration than in those with normal ε L (67 ± 7% vs 71 ± 7%, respectively, P = .002). Of note, the conventional diastolic functional parameters were not different in patients with and without ε L alteration.
Determinants of LV Geometry at Baseline
Multivariate analyses were performed to evaluate the independent association between ε L alteration and LV geometric parameters ( Table 2 ).
|Age (for a 1-y increase)||−0.22||.03||—||—|
|Heart rate (for a 1 beat/min increase)||−0.14||.03||—||—|
|HbA 1c (for a 1% increase)||−1.5||.008||−1.2||.02|
|ε L alteration||5.7||.001||5.0||.006|
|Peripheral artery disease||0.03||.07||—||—|
|ACE inhibitors or ARBs||0.04||<.001||—||—|
|LVEF (for a 1% increase)||0.02||.018||0.003||.005|
|ε L alteration||0.03||.008||0.03||.025|