Circulation Journal
Online ISSN : 1347-4820
Print ISSN : 1346-9843
ISSN-L : 1346-9843
Heart Failure
Impaired Mechanics of Left Ventriculo-Atrial Coupling in Patients With Diabetic Nephropathy
Yasuhide MochizukiHidekazu TanakaKensuke MatsumotoHiroyuki SanoHiroyuki ShimouraJunichi OokaTakuma SawaKeiko Ryo-KoriyamaYushi HirotaWataru OgawaKen-ichi Hirata
Author information

2016 Volume 80 Issue 9 Pages 1957-1964


Background: Diabetic nephropathy is independently associated with longitudinal systolic dysfunction of the left ventricle (LV) in asymptomatic diabetes mellitus (DM) patients with preserved LV ejection fraction (LVEF). However, the effect of diabetic nephropathy on left atrial (LA) function remains unknown.

Methods and Results: We studied 198 asymptomatic DM patients (LVEF ≥50%). Diabetic nephropathy was defined as a protein level higher than for micro-albuminuria. LV global longitudinal strain (GLS) and LA strain were analyzed by 2D speckle-tracking; 69 age-, sex-, and LVEF-matched controls were also studied. GLS and LA strain in systole (LAS-s) decreased significantly from normal controls to DM patients without (n=137) and with nephropathy (n=61), in that order. Furthermore, GLS, LAS-s, and LA strain in late diastole (LAS-a) were significantly lower in DM patients with macro-albuminuria (n=19) than in those with micro-albuminuria (n=42). Although 1 multivariate regression analysis identified albuminuria as an independent determinative factor of LAS-s among other relevant clinical background factors (β=−0.16, P=0.002), another multivariate regression model for LAS-s+GLS (β=0.40, P<0.001) showed that albuminuria was not a significant factor (β=−0.02, P=0.68). Similarly, another multivariate regression model including GLS (β=0.32, P<0.001) demonstrated that clinical features relevant for LAS-a, except for age, were not independent determinants of LAS-a.

Conclusions: The cross-linked association of LA strain with GLS and albuminuria may be important for better understanding the development of diabetic cardiomyopathy. (Circ J 2016; 80: 1957–1964)

Diabetes mellitus (DM) is considered to be one of the most significant causes of left atrial (LA) remodeling as well as left ventricle (LV) longitudinal systolic myocardial dysfunction (LVSD), which can increase risk of development of heart failure with preserved ejection fraction (EF) or atrial fibrillation (AF) even in asymptomatic DM patients with preserved LVEF.1,2 In fact, it has been reported by several epidemiological researchers that DM duration and poor glycemic control contribute to the morbidity of AF,35 and this relationship has been further substantiated by pathological evidence of LA fibrosis and remodeling in diabetic rats.6,7 The recent use of speckle-tracking strain imaging has made it possible to quantify detailed aspects of LA function such as reservoir, conduit and contractile function even before structural remodeling occurs.8 In addition, DM-related complications have been found to play an important role in the development of diabetic cardiomyopathy,912 and diabetic nephropathy has been identified as independently associated with LVSD in asymptomatic DM patients with preserved LVEF.1315 We recently focused our research especially on the interaction of albuminuria and LVSD in such patients,16 and as a result consider albuminuria to be the primary mediator of myocardial fibrosis by suppression of collagen turnover via impaired crosslinking of collagen. However, the effect of albuminuria on LA function in asymptomatic DM patients with preserved LVEF remains unknown.

Accordingly, the aim of this study was to investigate the effect of diabetic nephropathy on LA function, as assessed by 2D speckle-tracking strain echocardiography in asymptomatic DM patients with preserved LVEF. We also evaluated the relationship between LVSD and LA function in these patients.


Study Population

We enrolled a total of 204 consecutive DM patients admitted to Kobe University Hospital between July 2013 and September 2015 who did not meet any of the following exclusion criteria: (1) history of coronary artery disease (CAD); (2) LVEF <50%; (3) previous history of open-heart surgery or congenital heart disease; (4) severe renal dysfunction defined as glomerular filtration rate (GFR) <30 ml/min/1.73 m2; (5) uncontrolled hypertension >180/100 mmHg; (6) more than moderate valvular heart disease; (7) AF or flutter; and (8) left bundle branch block. All enrolled patients underwent an exercise stress screening test for CAD such as a treadmill exercise or stress myocardial perfusion scintigraphy during their hospitalization, and patients with an ischemic response were excluded from study entry. Moreover, strain analyses were excluded for 6 patients (3%) who were rated ineligible on the basis of echocardiographic images. Accordingly, 198 DM patients were enrolled in the final study group. For baseline comparison, a normal control group consisting of 69 age-, sex- and LVEF-matched subjects without a history of hypertension, DM, malignant disease or other cardiovascular disease and none of whom showed abnormal ECG findings. This study protocol was approved by the Ethics Board of Kobe University and each patient gave informed consent for this study before participation.


Within less than 2 weeks of admission, all participants underwent resting transthoracic echocardiography using a commercially available system (Vivid E9; General Electric Medical Systems, Milwaukee, WI, USA) with a 3.5-MHz transducer.17 Digital routine grayscale cine-loops were obtained for 3 consecutive heartbeats from standard parasternal long-axis and 3 apical views. Sector width (11–17 cm) and frame rate (60–80 frames/s) were optimized with particular care for each subject to obtain complete LV and LA myocardial visualization. Raw data thus obtained were then transferred to dedicated software for subsequent speckle-tracking analysis. Conventional LV and LA parameters were measured according to the current guidelines of the American Society of Echocardiography/European Association of Cardiovascular Imaging.18 Specifically, LV and LA volumes and LVEF were obtained by using the modified biplane Simpson’s method, and LA volume was then normalized to body surface area. LV stroke volume was calculated by multiplying the velocity-time integral, assessed by means of pulsed-wave Doppler positioned at the LV outflow tract, by its area. Pulsed-wave Doppler from the apical long- or 4-chamber view was used to measure early diastolic (E) and atrial wave (A) velocities, E-wave deceleration time. Spectral pulsed-wave Doppler-derived early diastolic velocity (E’) was obtained from the septal mitral annulus.

LV and LA Speckle-Tracking Analysis

Two-dimensional speckle-tracking strain analysis was semi-automatically performed with dedicated software (EchoPAC version 113; General Electric Medical Systems). Briefly, the first region of interest was manually traced on the endocardium of LV and LA at the end-systole phase with the point-and-click approach. The second larger region of interest was then generated outside and carefully adjusted near the epicardium. Finally, 6 strain segments and corresponding time-strain curves were generated. We used the onset point of the QRS complex as a reference for both LV and LA strain analysis. Global longitudinal strain (GLS) was then determined as the averaged peak strain from standard 3-apical views in accordance with current guidelines18 (Figure 1). LA reservoir (LAS-s), passive emptying (LAS-e), and contractile (LAS-a) functions were defined as the average value of global LA longitudinal strain at peak systole, early- and end-diastole in 2- and 4-chamber views as previously described19 (Figure 2). All strain values were expressed as an absolute value in this study to avoid confusion about magnitude relationships, as is recommended in the current guidelines.18

Figure 1.

An example of global longitudinal strain evaluated by 2D speckle tracking imaging from 3 standard apical views and 18 corresponding longitudinal strain curves and using onset of the QRS complex as the starting point of strain.

Figure 2.

2D speckle tracking imaging-derived left atrial (LA) strain analysis from apical 2- and 4-chamber views using software for LV strain analysis. Each average of the peak global strain curve (white) in systole, early diastole, and end-diastole was determined as LA strain in systole (LAS-s), early diastole (LAS-e) and late diastole (LAS-a).

Clinical Data for DM Patients

Fasting blood glucose, hemoglobin A1c, estimated GFR and lipid profile were examined on the day following admission. Dyslipidemia was defined as fasting low-density lipoprotein ≥140 mg/dl, or current use of antidyslipidemia drugs.20 Blood pressure (BP) and heart rate were obtained simultaneously with transthoracic echocardiography. Hypertension was then defined as systolic BP ≥140 mmHg or diastolic BP ≥90 mmHg, or current treatment with antihypertensive agents.21 Albuminuria was quantitated by means of 24-h urine collection, micro-albuminuria was defined as urinary albumin excretion (UAE) within >20 μg/min and ≤199 μg/min, and macro-albuminuria as UAE ≥200 μg/min on the basis of guidelines concerning diabetic nephropathy.22 We then defined diabetic nephropathy as a protein level higher than that for micro-albuminuria for classification of the DM patients into 2 groups: one without and the other with nephropathy. All subjects were subjected to a thorough interview and methodical and comprehensive physical examination by experienced diabetologists. Diabetic retinopathy was defined as more serious than mild non-proliferative diabetic retinopathy or as evidence of treatment by laser photocoagulation on the basis of evaluation of retinal photographs by experienced ophthalmologists.23

Statistical Analysis

SPSS version 16.0 (SPSS Inc, Chicago, IL, USA) and MedCalc version 14.12.0 (MedCalc Software, Mariakerke, Belgium) were used for all steps of statistical analyses. Statistical significance was defined as P<0.05. Continuous variables are expressed as mean values and standard deviation for normally distributed data and median and as an interquartile range for non-normally distributed data. Continuous parameters among subgroups were evaluated by 1-way analysis of variance, and the Tukey-Kramer test was then used for posthoc analysis. Categorical variables are expressed as frequencies and percentages. Proportional differences were evaluated with Fisher’s exact test and Chi-square test. Relationships between 2 variables were analyzed by linear regression and expressed as Pearson correlation coefficients. A multivariate regression analysis for detecting LAS-s and LAS-a was performed with variables relevant only for patients’ clinical characteristics that showed significant correlation in the univariate analysis. Another multivariate model with the addition of GLS was designed because GLS is closely associated with albuminuria and LA longitudinal extension.16,24 The inter- and intra-observer variabilities for GLS and LA assessments were determined by intra-class correlation coefficient from 20 randomly selected patients.


Baseline Characteristics

As previously mentioned, the DM patients were divided into 2 groups: those without (n=137) and with a protein level higher than that for micro-albuminuria (n=61) (the latter forming the DM nephropathy group). In addition, the nephropathy group was further divided into 2 groups: patients with micro-nephropathy (n=42) and those with macro-nephropathy (n=19). Table 1 shows the baseline clinical characteristics of the normal subjects and the 2 groups of DM patients. Age and sex ratios were similar for the 3 groups. DM patients with nephropathy had DM of longer duration and more metabolic abnormalities related to DM compared with DM patients without nephropathy. However, parameters of the blood examination representing recent glycemic condition were similar for the 2 groups with DM. Conventional echocardiographic parameters are also shown in Table 1. LV diastolic dysfunction was the most advanced for DM patients with nephropathy, while LVEF and LV volumes were similar for the 3 groups. As expected, LA volumes and LV mass were significantly larger in the DM nephropathy group than in the other 2 groups.

Table 1. Baseline Clinical Characteristics of Asymptomatic Diabetic Patients
  Normal controls
DM without
nephropathy (n=137)
DM with
nephropathy (n=61)
Clinical data
 Age, years 52±16 55±15 58±16
 Female, n (%) 43 (62) 77 (56) 30 (49)
 BMI, kg/m2 21±3 24±4 26±6*
 Systolic BP, mmHg 120±15 121±16 135±26*
 Heart rate, beats/min 65±10 68±12 71±10
 QRS duration, ms 88±9 90±14 92±13
 DM duration, years 10±9 14±10
 Type 2 DM, n (%) 85 (62) 53 (87)
 Hypertension, n (%) 55 (40) 40 (66)
 Dyslipidemia, n (%) 75 (55) 36 (59)
 Retinopathy, n (%) 37 (27) 36 (59)
Blood and urinary tests
 HbA1c, % 8.6±2.2 9.0±2.0
 Glycoalbumin, % 24.3±9.1 23.9±8.7
 LDL, mg/dl 104±33 110±36
 Triglyceride, mg/dl 124±74 157±96
 eGFR, ml/min/1.73 m2 80±22 69±30
 Albuminuria, mg/day 7 (3–14) 104 (42–372)
 CCB, n (%) 29 (21) 26 (43)
 ACEI/ARB, n (%) 46 (34) 33 (54)
 β-blocker, n (%) 7 (5) 9 (15)
 Statin, n (%) 51 (37) 31 (51)
 Insulin, n (%) 87 (64) 42 (69)
 DPP-4I, n (%) 46 (34) 30 (49)
 GLP-1RA, n (%) 10 (7) 4 (7)
 Sulfonylurea, n (%) 20 (15) 14 (23)
 α-GI, n (%) 27 (20) 10 (16)
 Thiazolidine, n (%) 9 (7) 8 (13)
 Metformin, n (%) 57 (42) 22 (36)
 LA volume index, ml/m2 26±8 28±7 33±10*
 LV mass index, g/m2 72±17 71±17 86±22*
 LV end-systolic volume, ml 27±10 25±9 28±12
 LV end-diastolic volume, ml 77±23 74±20 80±23
 LVEF, % 66±5 66±4 65±5
 Stroke volume, ml 69±15 63±12 66±13
 E/A 1.2±0.5 1.0±0.5 0.9±0.4
 DT 179±35 194±50 197±52
 IVRT 67±16 83±22 88±18
 E’ 9.7±3.1 7.2±2.7 6.0±2.1*
 E/E’ 7.8±2.3 9.5±3.0 12.1±5.2*

*Significant difference vs. normal controls and DM without nephropathy (P<0.05); significant difference vs. normal controls (P<0.05); significant difference vs. DM without nephropathy (P<0.05). α-GI, α-glucosidase inhibitor; A, peak late diastolic mitral flow velocity; ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; BMI, body mass index; BP, blood pressure; CCB, calcium-channel blocker; DM, diabetes mellitus; DPP-4I, dipeptidyl peptidase-4 inhibitor; DT, E-wave deceleration time; E, peak early diastolic mitral flow velocity; E’, Spectral pulsed-wave Doppler-derived early diastolic velocity from the septal mitral annulus; EF, ejection fraction; eGFR, estimated glomerular filtration rate; GLP-1RA, glucagon-like peptide-1receptor agonist; IVRT, isovolumic relaxation time; LDL, low-density lipoprotein; LV, left ventricular.

LA Strain and GLS in Stages of Diabetic Nephropathy

LAS-e did not show significant differences among the 3 groups or even between DM patients with micro- and macro-nephropathy. Figure 3 clearly identifies differences in GLS, LAS-s and LAS-a for various stages of diabetic nephropathy. GLS and LAS-s decreased significantly from normal controls to DM patients without and with nephropathy, in that order, and were also significantly lower for DM patients with macro-albuminuria than for those with micro-albuminuria. Although LAS-a did not differ significantly within the 3 groups, it was smaller in DM patients with macro-albuminuria than in those with micro-albuminuria.

Figure 3.

(A) Bar graphs of GLS, LAS-s and LAS-a for normal subjects, and for DM patients with and without nephropathy. A stepwise decrease in GLS and LAS-s was observed among the groups, and a similar finding was obtained for LAS-a, but the difference was not statistically significant. (B) Bar graphs of GLS, LAS-s and LAS-a for DM patients with micro-albuminuria and macro-albuminuria. GLS, LAS-s and LAS-a for the former were significantly higher than for the latter. DM, diabetes mellitus; GLS, global longitudinal strain; LAS-s, LA strain in systole; LAS-e, LA strain in early diastole; LA strain in LAS-a late diastole; NS, not significant.

Clinical Determinants of LA Reservoir and Contractile Function

Table 2 summarizes the results of univariate and multivariate regression analyses for LAS-s and LAS-a using only the clinical characteristics of the DM patients (n=198). Multivariate regression analysis following individual univariate regression analyses showed that age, body mass index (BMI), triglycerides and log-transferred albuminuria were significant determinative factors for LAS-s, while another multivariate regression for LAS-a revealed that only age and BMI were independent determinative factors.

Table 2. Univariate and Multivariate Regression Analyses for LAS-s and LAS-a in Asymptomatic Diabetic Patients
Dependent variables LAS-s LAS-a
Univariate Multivariate Univariate Multivariate
β P value β P value β P value β P value
Age −0.57 <0.001 −0.58 <0.001 0.18 0.013 0.17 0.014
Sex (female) 0.11 0.133     0.04 0.541    
DM duration −0.16 0.028     0.03 0.682    
BMI −0.34 <0.001 −0.27 <0.001 −0.15 0.032 −0.15 0.035
Type 2 DM −0.44 <0.001     0.08 0.241    
Systolic BP −0.37 <0.001     0.02 0.739    
Triglyceride −0.25 <0.001 −0.15 0.007 −0.12 0.101    
Hemoglobin A1c −0.01 0.926     −0.07 0.298    
Albuminuria −0.26 <0.001 −0.16 0.002 −0.12 0.081    
Retinopathy −0.17 0.018     0.01 0.890    

β, standardized β; LAS-a, left atrial strain in late diastole; LAS-s, left atrial strain in systole. Other abbreviations as in Table 1.

Association Between Clinical Background, LA Strain and GLS in DM Patients

GLS showed significant correlations with albuminuria (r=−0.45, P<0.001) and BMI (r=−0.35, P<0.001), as well as LAS-s (r=0.54, P<0.001) and LAS-a (r=0.32, P<0.001). LAS-s, on the other hand, showed a weak yet significant correlation with albuminuria (r=−0.16, P=0.002), while LAS-a showed a similar weak correlation with BMI (r=−0.15, P=0.035). Therefore, when GLS was added as a covariate to these correlations, the partial correlations of LAS-s with albuminuria and LAS-a with BMI did not remain significant. As shown in Table 3, 2 multivariate regression models for LAS-s and LAS-a were performed in a similar manner, one of which was constructed only with independent clinical background factors selected from data obtained by preceding multivariate regressions (“clinical only” model), and the other was a modified model with GLS added as a strong confounding factor. This modified multivariate model clearly demonstrated that powerful clinical background factors associated with LA strains barely remained significant when GLS was added to the model.

Table 3. Multivariate Regression Analyses for LAS-s and LAS-a in Asymptomatic Diabetic Patients
Dependent variables LAS-s LAS-a
Clinical only* Clinical+GLS** Clinical only* Clinical+GLS**
β P value β P value β P value β P value
Age −0.58 <0.001 −0.54 <0.001 0.17 0.014 0.20 0.003
BMI −0.27 <0.001 −0.19 <0.001 −0.15 0.035 −0.03 0.633
Triglyceride −0.15 0.007 −0.06 0.226        
Albuminuria −0.16 0.002 −0.02 0.678        
GLS     0.40 <0.001     0.32 <0.001

*Enter variable if P<0.05 by stepwise selection; **enter covariates forcibly. GLS, global longitudinal strain. Other abbreviations as in Tables 1,2.

Reproducibility of GLS and LA Strain

The intra- and inter-observer variabilities evaluated by means of intra-class correlation coefficient were 0.948 (95% confidence interval (CI): 0.855–0.982) and 0.921 (95% CI: 0.775–0.973) for GLS, 0.975 (95% CI: 0.932–0.991) and 0.959 (95% CI: 0.890–0.985) for LAS-s, and 0.930 (95% CI: 0.814–0.974) and 0.895 (95% CI: 0.720–0.961) for LAS-a.


The main findings of our study were as follows. (1) LA reservoir function of DM patients as assessed by 2D speckle-tracking strain was significantly reduced compared with a normal healthy population, and was even more diminished in cases of diabetic nephropathy linearly associated with albuminuria. LA contractile function was also diminished, especially in patients with macro-albuminuria, and obesity was an independent clinical feature associated with LAS-a. (2) GLS also showed a stepwise decrease in line with the degree of diabetic nephropathy, which correlated significantly with both LA strain and albuminuria. (3) Multivariate analysis including clinical factors for detecting LAS-s and LAS-a revealed that adjustment with GLS eliminated the significant associations of albuminuria with LAS-s and of BMI with LAS-a.

Speckle Tracking-Derived LA Function in DM Patients

Two-dimensional speckle-tracking imaging is a well-known method for the evaluation of LA reservoir function, which reportedly is a prognostic factor of cardiovascular diseases. Moreover, a decrease in LA reservoir function has been detected only in asymptomatic patients with cardiovascular risks such as hypertension and DM.19,2528 We investigated the association between LA strain and DM in the clinical setting, especially the stage of diabetic nephropathy, because worsening of nephropathy reportedly has a strong relationship with LV systolic and diastolic dysfunction even in asymptomatic DM patients.13,14 This study identified the stage of diabetic nephropathy as a factor associated with LA reservoir function (LAS-s) and contractile function (LAS-a). We also detected an independent and negative association of BMI with LA contractile function (LAS-a). In addition, LAS-a was significantly more reduced in patients with macro-albuminuria than in those with micro-albuminuria. The MONICA study demonstrated that the main cause of LA remodeling was obesity rather than age and hypertension.29 Additionally, it was reported that GLS in obese patients was significantly reduced in line with the degree of metabolic syndrome, especially in patients with central obesity.30 Accordingly, our finding from the multivariate analysis as shown in Table 3 suggested that a reduction in GLS through obesity and progression of diabetic nephropathy led to loss of LAS-a, while aging played a role in LA compensatory regulation.

Pathophysiological Relation Between LA Dysfunction and LVSD

Morris et al reported that symptomatic patients with heart failure and preserved EF had significantly lower GLS and LA strain than did their asymptomatic counterparts.31 They also showed that GLS closely correlated with both systolic and diastolic LA strains. Impairment of GLS led to a decrease in the force of drawing the LV basal plane in systole, and this mechanical link to insufficient pulling of the left atrium into the apex. Under such circumstances, poor LA compliance because of atrial fibrosis accelerates the reduction of LA passive extension, resulting in loss of LA reserve. Because the mechanics of LA contraction reciprocate against LV pressure as an afterload for LA in the pre-atrial contraction period, it plays in the same way an important role for smooth passive stretching in LV end-diastole, whereas a rigid LV characterized as indicating a reduced GLS may lower the LA contractile functional evaluation.3234 Moreover, loss of LA reserve driven by a reduction of GLS in systole may reduce the LA emptying function as a length-tension relationship.34 It therefore follows that LVSD is closely associated with LA reservoir and emptying dysfunction.

Clinical Implications of LA Strain in DM Patients

We previously demonstrated that diabetic nephropathy is the most detrimental factor for reduced GLS.16 In addition, Jensen et al recently found in their Thousand & 1 study that GLS in patients with type 1 DM but without albuminuria was equivalent to that in normal healthy controls.35 Diabetic nephropathy can be viewed as the pathophysiological mirror of an uncontrolled glycemic state and attendant complications such as hypertension and other metabolic changes. In DM patients with nephropathy, activation of the renin-angiotensin-aldosterone system, which leads to an increase in advanced glycation endproducts, as well as hyperinsulinemia and dysregulation of extracellular matrix, which induces intense fibrosis, at the myocardial cellular level.10 To evaluate LA function in DM patients, even at the preclinical stage, is crucially important. Although LA function of DM patients is greatly influenced by LV longitudinal shortening, assessment of LA function may be an additional factor that facilitates earlier detection of subclinical cardiac dysfunction in DM patients with preserved LVEF and diabetic nephropathy.

Study Limitations

This study was a single-center cross-sectional study, so further longitudinal-based cohort studies are required to validate our results. Another limitation is that long-term clinical outcome data such as that for associations of LA strain were not part of the study. Because the vendor’s speckle-tracking software for the LA has not yet been developed, we used a speckle-tracking program for LV strain to assess LA strain because other investigators have used the same speckle-tracking program for LV strain to assess LA strain, and determined that the feasibility and reproducibility of LA strain were acceptable.19,24,25


LA reservoir and contractile function are more impaired in DM patients with albuminuria than in those without albuminuria. Because the effect of albuminuria on LA function is strongly mediated by LV longitudinal systolic function, it may contribute to a better understanding of the development of diabetic cardiomyopathy.