Diabetes and diastolic function: Stiffness and relaxation from transmitral flow

Ultrasound in Med. & Biol., Vol. 31, No. 12, pp. 1589 –1596, 2005 Copyright © 2005 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/05/$–see front matter

doi:10.1016/j.ultrasmedbio.2005.07.015

● Original Contribution DIABETES AND DIASTOLIC FUNCTION: STIFFNESS AND RELAXATION FROM TRANSMITRAL FLOW MATT M. RIORDAN, CHARLES S. CHUNG and SÁNDOR J. KOVÁCS Cardiovascular Biophysics Laboratory, Cardiovascular Division, Department of Internal Medicine, Washington University School of Medicine, St. Louis, MO USA (Received 11 April 2005; revised 1 July 2005; in final form 7 July 2005)

Abstract—To characterize the mechanism by which diabetes affects the heart in diabetic (n ⴝ 15) and age-matched control subjects (n ⴝ 15), we quantified and compared diastolic function (DF) in terms of chamber stiffness and viscosity/relaxation by analyzing Doppler E- and Eⴕ-waves and simultaneous (high-fidelity) hemodynamic data. We compared ␶, standard Doppler indexes and indexes of stiffness and viscosity/relaxation computed via the parameterized diastolic filling (PDF) formalism. Three PDF parameters uniquely characterize each E-wave in terms of load (xo), viscoelasticity or viscosity/relaxation (c) and stiffness (k). Significant differences for c (p ⴝ 0.00004), the peak atrioventricular pressure gradient (kxo) (p ⴝ 0.02) and the stored elastic energy available for early filling (1/2kx2o) (p ⴝ 0.04) were found. The only conventional index attaining significance was E-wave acceleration time (p ⴝ 0.007). Neither time constant of isovolumic relaxation (␶) nor E-wave deceleration time, Eⴕ, k or xo differentiated between groups. We conclude that PDF based DF assessment differentiates between diabetic and nondiabetic controls better than conventional echo- or cath-based indexes. Our results in humans agree with published results from animal studies. We conclude that diabetes affects the heart via a quantifiable increase in chamber viscoelasticity (c) rather than an increase in chamber stiffness (k) and that phenotypic characterization of diabetic cardiomyopathy is facilitated by DF assessment via the PDF formalism. (E-mail: [email protected]) © 2005 World Federation for Ultrasound in Medicine & Biology. Key Words: Diastole, Echocardiography, Diabetes, Physiological modeling, Diabetic cardiomyopathy, E-wave, Deceleration time.

as a contributing cause to cardiac dysfunction (Zile et al. 2004). Abnormalities in diastolic function (DF) have been shown to precede abnormalities in systolic function in diabetic cardiomyopathy (Raev 1994; Lo et al. 1995). This reality underscores the importance of early diagnosis and initiation of therapy to prevent dysfunction before overall chamber function is further as well as irreversibly impaired.

INTRODUCTION Diabetic patients live with an increased risk of morbidity and mortality due to associated cardiovascular complications, notably hypertension, coronary artery disease (CAD) and microvascular disease (Aronson et al. 1997). Despite the likely contribution of such complications to cardiac dysfunction, diabetic cardiomyopathy can exist in their absence (Paillole et al. 1989; Regan and Weisse 1992; Shehadeh and Regan 1995). Previous studies have demonstrated an alteration of the molecular and cellular properties (glycosylation of proteins and collagen, advanced glycosylation end-products [AGEs]) of the diabetic heart. These changes, which include concomitant abnormalities of lipid metabolism (Rodrigues et al. 1998), promote ventricular stiffness, which is often cited

Conventional indexes derived from transmitral flow Noninvasive assessment of DF is commonly achieved by 2-D pulsed Doppler echocardiography (Appleton et al. 2000; Naqvi 2003). Analysis of transmitral flow patterns is generally based on manually determined geometric features of the flow pattern, such as the ratio of the peak velocities of the E- and A-wave (E/A) and deceleration time (DT), which have been correlated with presence of dysfunction (Andersen et al. 2003). While these indexes are often indicative of dysfunction, their utility in characterizing relaxation and

Address correspondence to: Sándor J. Kovács, Ph.D., M.D., Cardiovascular Biophysics Laboratory, Washington University Medical Center, Box 8086, 660 South Euclid Ave., St. Louis, MO 63110 USA. E-mail: [email protected] 1589

1590

Ultrasound in Medicine and Biology

stiffness is limited (Cosson and Kevorkian 2003). Furthermore, since only one or two points of the E-wave contour are utilized for their computation, the information contained in the entire E-wave contour is neglected (Hall et al. 1996; Appleton et al. 2000). Sensitivity and specificity may be limited, as exemplified by advanced stages of diastolic dysfunction (DD), when E/A may become “pseudonormalized” (E/A⬎1), effectively masking the diagnosis of DD (Appleton et al. 1988). Finally, the dependence of the conventional indexes on certain variables and echocardiography machine settings is incompletely understood and, therefore, may represent an additional source that confounds the reliability of DF assessment (Hall et al. 1996). Diastolic function can also be characterized echocardiographically, based on the intraventricular flow (Vp) and gradient that must be generated by the mechanical suction attribute of the LV. (Kovács et al. 1987; Courtois et al. 1988; Thomas et al. 1991). The resultant effect of the computed intraventricular gradient so characterized via color M-mode Doppler is the contour of the transmitral Doppler E-wave. PDF formalism based indexes derived from transmitral flow The primary advantage of the parameterized diastolic filling (PDF) formalism in characterizing DF, besides its observer-independent aspects, is that it is “predictive” rather than “accommodative” in characterizing the E-wave contour (Lipton 2005). Furthermore, its parameters have conceptually and experimentally well-established physiological analogs. By modeling filling of the left ventricle (LV) kinematically (based on how things move) in analogy to the motion of a linear, damped simple harmonic oscillator (SHO), the stiffness, damping/viscosity and initial displacement (load) of the system can be determined uniquely on a beat-by-beat basis. Therefore, by using the E- (and A-) wave contour as input to the formalism, the PDF approach solves the “inverse problem of diastole” and generates unique parameters for each E-wave analyzed (Kovács et al. 2000). When applied in the clinical arena, the PDF formalism has successfully differentiated hypertensive hearts from controls (Kovács et al. 1997), provided the most sensitive and specific index of 1-year mortality for hospitalized, elderly patients with congestive heart failure (Rich et al. 1999), elucidated the role of chamber stiffness as a determinant of exercise in subjects with HF (Meyer et al. 2004), provided an improved index for determination of the maximum atrioventricular pressure gradient that generates the E-wave (Bauman et al. 2004) and elucidated the relationship between the ratio of the peak velocities of the E-wave and E⬘-wave (E/E⬘) and DF in terms of left ventricular end-diastolic pressure (LVEDP) in causal rather than correlative terms (Lisauskas et al. 2001).

Volume 31, Number 12, 2005

Previously, the PDF approach to DF characterization has allowed differentiation of diabetic rat hearts from control hearts (Dent et al. 2001). Therefore, the main objective of the present study was to determine if E-wave analysis via the PDF formalism provides DF indexes that distinguish DF in normotensive diabetic humans and an age-matched, normotensive nondiabetic control group. In addition, if index based differences were established between groups, we sought to characterize the physiological basis of impaired left ventricular filling in terms of stiffness vs. relaxation/viscosity. Conventional echocardiographic indexes (acceleration time [AT], DT, E/A, E/E⬘, E⬘/A⬘, isovolumic relaxation time [IVRT]) and simultaneously recorded hemodynamic indexes (LVEDP, dP/dtmax, dP/dtmin, ␶) were calculated for each group as well to determine their ability to distinguish between diabetics and nondiabetic controls. METHODS Patient selection A sample of 30 subjects aged between 44 and 71 was selected from an existing database (Lisauskas et al. 2001) of simultaneous high fidelity (Millar) ventricular pressure, Doppler echocardiographic recordings of transmitral flow and Doppler tissue recordings of motion of the lateral mitral annulus. Inclusion criteria for the study was that all subjects be normotensive at the time of data acquisition (naturally, or because of antihypertensive therapy), in normal sinus rhythm, have clearly discernible E-waves, have normal valvular function and possess diabetes as an established previous diagnosis as part of their documented medical history. In addition, individuals with comorbidities including, but not limited to, previous myocardial infarction (MI), peripheral vascular disease, bundle branch block, ischemia, cardiomyopathy, congestive heart failure or renal insufficiency were excluded. All subjects had a normal ejection fraction (EF) (EF ⬎ 50%) by ventriculography. The two groups were composed of 15 individuals with a prior diagnosis of diabetes based on their medical history and 15 individuals with no history of diabetes (the nondiabetic control group). Catheterization in these subjects was performed because the patients were referred by their cardiologist on the basis of suspected coronary artery disease (CAD). Diabetic subjects in the database meeting the inclusion criteria and not possessing any exclusion criteria were selected for this study. The nondiabetic controls were chosen with the intent of matching, as closely as possible, the demographics of the diabetic group, including race, gender, age, height, weight and history of smoking. Furthermore, an equal number of subjects in each group (seven) had presence of angiographically diagnosed CAD. These clinical attributes are summarized in Table

Diastolic function in diabetes ● M. M. RIORDAN et al.

1. All subjects gave informed consent according to WUMC Human Studies requirements before data acquisition. Data acquisition The methodology employed has been previously described (Lisauskas et al. 2001). Briefly, immediately before cardiac catheterization, a full 2-D/Doppler examination was performed in the catheterization laboratory. Transmitral flow velocity acquisition was performed simultaneous with LV pressure recording as previously described (Lisauskas et al. 2001). Although some subjects had narrowing of the coronary arteries (⬎50% narrowing), no subject among the cohort had active ischemia at the time of data acquisition or prior MI. After advancement of the micromanometric 6F pigtail catheter (Millar Instruments, Houston, TX, USA, Model SPC560) into the LV, transmitral Doppler images were obtained by an experienced sonographer using a standard clinical imaging system (Acuson, Mountain View, CA, USA). With the patient supine, apical four-chamber views were obtained with the sample volume gated at 1.5 to 2.5 mm and placed at the tips of the mitral valve leaflets with the insonification direction orthogonal to the valve plane using color Doppler as a guide. The wall filter was set at 125 or 250 Hz, the baseline adjusted to take advantage of the full height of the CRT display and the velocity scale adjusted to exploit the dynamic range of the output without aliasing. Simultaneous limb lead II ECG was displayed on all images. Images were captured simultaneously with LV pressure data and recorded continuously via VHS or magneto-optical disk. Doppler tissue imaging (DTI) was performed in similar fashion with the sample volume gated between 2.5 and 5 mm and positioned at the lateral side of the mitral annulus. Images were digitized off-line via a dedicated custom video capture station. Doppler analysis For each subject, five transmitral Doppler E- and A-waves were selected, clipped and converted to 8-bit Table 1. Clinical variables

Gender (men) Minorities Age (y) Height (cm) Weight (kg) Ejection fraction (%) Mean arterial pressure (mmHg) Heart rate (bpm)

Control group (n ⫽ 15)

Diabetic group (n ⫽ 15)

10 3 54 ⫾ 9 170 ⫾ 13 82 ⫾ 26 71 ⫾ 7 94.9 ⫾ 15.0 66 ⫾ 12

10 4 55 ⫾ 11 175 ⫾ 10 94 ⫾ 24 68 ⫾ 11 101.9 ⫾ 14.4 72 ⫾ 10

Data are presented as mean value ⫾ SD. No significant differences exist in these variables between groups.

1591

grayscale images with Paint Shop Pro 7 (Jasc Software, Minnetonka, MN, USA). E-waves were fit using modelbased image processing (MBIP) according to the previously validated PDF method (Kovács et al. 1987; Hall and Kovács 1994). In brief, the MBIP minimizes the nonlinear least-squares error between the model-predicted velocity and actual E-wave contour via the Levenberg-Marquardt algorithm (Press et al. 1986). The program uses the maximum velocity envelope of the raw image (E-wave) as input and generates three parameters, c, k and xo, as output. The parameters are defined by the equation of motion for a SHO: m(d2x ⁄ dt2) ⫹ c(dx ⁄ dt) ⫹ kx ⫽ 0

(1)

where c (g · cm/s) and k (g · cm/s2) denote the damping constant and spring constant of the system, respectively. The parameter xo (cm) accounts for load and represents the initial displacement of the spring before motion and corresponds to the elastic strain stored in the myocardium and surrounding structures available at mitral valve opening that facilitates mechanical recoil (Kovács et al. 2000). Corresponding to no transmitral flow before valve opening, the initial velocity (dx/dt) of the system is zero. The inertial term m (g) is normalized to 1 to enable the computation of c and k per unit mass. Additional SHO derived indexes include kxo, the peak force in the spring, corresponding to the peak atrioventricular pressure gradient that generates the E-wave (Kovács et al. 1987; Bauman et al. 2004) and 1/2kx2o, the stored potential elastic energy capable of generating recoil (Kovács et al. 1987). For each subject, the average of c, k and xo, as well as kxo, 1/2kx2o and ␤ ⫽ (c2⫺4mk), which assesses the relative contribution between damping (c) and recoil (k) as determinants of E-wave, were calculated from the five selected beats. Conventional echocardiographic parameters that were measured include: AT, DT, peak E- and A-wave velocity (Epeak and Apeak, respectively), E/A and IVRT. These were calculated for each of the five E- and A-wave images for each subject and averaged. Peak early mitral annular velocity (E⬘) and peak atrial mitral annular velocity (A⬘), E/E⬘ and E⬘/A⬘ were calculated similarly from Doppler tissue recordings. As is the custom, these parameters were all measured manually, with DT calculated using standard triangle approximation for E-wave shape (Feigenbaum 1994). IVRT was measured on the Doppler tracing as the time between closure of the aortic valve and opening of the mitral valve as evidenced by the leading edge of the E-wave (Feigenbaum 1994). Hemodynamic analysis Mean arterial pressure (MAP) was determined from the aortic root pressure and LVEDP was determined

1592

Ultrasound in Medicine and Biology

from the LV pressure tracing (Bauman et al. 2004). Peak positive and peak negative rates of change of pressure, (dP/dtmax, dP/dtmin) and ␶ (Eucker et al. 2001) were calculated using the pressure “phase-plane” (i.e., the plot of the first derivative of pressure [dP/dt] vs. pressure [P]). ␶ was computed as the inverse of the slope of the line fit between dP/dtmin and the point on the plot corresponding to mitral valve opening (Eucker et al. 2001). Statistical analysis All data are displayed as mean ⫾ standard deviation. Statistical differences between the entire diabetic and nondiabetic control group for PDF parameters, conventional echocardiographic parameters and hemodynamic parameters were determined by two-tailed analysis of variance (ANOVA). For each parameter, the null hypothesis tested was that the mean value for individuals without diabetes is equal to the mean value for diabetic individuals. All statistical calculations were performed in Microsoft Excel 97 (Microsoft Corporation, Redmond, WA, USA) and statistical significance was defined at the p ⬍ 0.05 level. RESULTS The groups were well matched, and no statistically significant differences were observed between the groups in any clinical category (Table 1). E-waves of the diabetic subjects revealed a significantly shorter AT, but no difference between diabetic and normal subjects was found for DT. The other measured conventional echocardiographic parameters of Epeak, Apeak, E/A, E⬘, A⬘,

Table 2. Conventional Doppler and hemodynamic parameters

Peak E (cm/s) Peak A (cm/s) E/A Acceleration Time (ms) Deceleration Time (ms) IVRT (ms) LVEDP (mmHg) dP/dtmax (mmHg/s)† dP/dtmin (mmHg/s)† tau (ms)‡ E’ (cm/s)* A’ (cm/s)* E/E’* E’/A’*

Control group (n ⫽ 15)

Diabetic group (n ⫽ 15)

67.1 ⫾ 10.0 63.2 ⫾ 20.5 1.16 ⫾ 0.41

70.2 ⫾ 20.4 66.1 ⫾ 13.4 1.06 ⫾ 0.19

P ⫽ 0.60 P ⫽ 0.65 P ⫽ 0.39

89.1 ⫾ 13.0

76.8 ⫾ 9.7

P ⫽ 0.007

199.1 ⫾ 39.1 49 ⫾ 18 14.9 ⫾ 6.4 1272 ⫾ 255 ⫺1368 ⫾ 335 50.3 ⫾ 16.9 15.3 ⫾ 4.2 16.7 ⫾ 4.3 4.70 ⫾ 1.61 1.15 ⫾ 0.22

210.3 ⫾ 44.6 52 ⫾ 13 14.3 ⫾ 5.5 1169 ⫾ 251 ⫺1631 ⫾ 271 51.0 ⫾ 16.0 17.2 ⫾ 2.7 15.1 ⫾ 3.1 4.08 ⫾ 1.44 1.03 ⫾ 1.41

Significance

P P P P P P P P P P

⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽ ⫽

0.47 0.59 0.77 0.51 0.06 0.91 0.24 0.66 0.38 0.28

* E’, A’, E/E’, and E’/A’ were only obtained for 11 control subjects and 9 diabetic subjects. † dP/dtmax and dP/dtmin were only obtained for 9 diabetic subjects. ‡ Tau was only obtained for 14 subjects in each group.

Volume 31, Number 12, 2005

Table 3. PDF Parameters and parameter derived indexes for all subjects

c (g/s) k (g/s2) x0 (cm) kx0 (dynes) 1/2kx20 (ergs) ␤ (g2/s2)

Control group (n ⫽ 15)

Diabetic group (n ⫽ 15)

Significance

17.7 ⫾ 4.7 183 ⫾ 54 10.7 ⫾ 2.1 1910 ⫾ 480 10400 ⫾ 1320 ⫺358 ⫾ 164

25.9 ⫾ 4.6 217 ⫾ 51 12.0 ⫾ 3.5 2590 ⫾ 880 16700 ⫾ 10400 ⫺158 ⫾ 277

P P P P P P

⫽ ⫽ ⫽ ⫽ ⫽ ⫽

0.00004 0.09 0.23 0.02 0.04 0.02

E/E⬘, E⬘/A⬘ and IVRT also showed no significant difference between groups (Table 2). Furthermore, MAP, dP/dtmax, dP/dtmin, LVEDP and ␶ did not differ significantly between the diabetic and nondiabetic control groups. PDF formalism derived indexes revealed statistically significant differences. Compared with nondiabetic controls, the diabetic group exhibited greater values for the parameter c. The diabetic group also exhibited greater values of kxo, 1/2kx2o and ␤, although no significant differences existed for k or xo individually (Table 3). Intra-observer reproducibility was assessed by twice fitting five representative E-waves of varying shapes from both the normal and diabetic group. The coefficients of variation (Bland and Altman 1996) for c, k and xo were determined to be 2.8%, 1.2% and 3.6%, respectively. The maximum variations in c, k and xo for these fits were 12.2%, 7.3% and 5.9%, respectively. The data shows a noticeable trend. It appears that the parameter c is greater in diabetic women than diabetic men. In fact, the lowest value of c in diabetic women is greater than the largest value of c for any subject, male or female, in the control group. Since our sample size is insufficient to form any gender-based conclusions, a gender-based study having sufficient statistical power to detect statistically reliable differences is planned. DISCUSSION In this study, the PDF formalism was used to characterize DF in normotensive age-matched diabetic subjects vs. nondiabetic controls. We characterized the effect of diabetes on DF through its influence on chamber stiffness and relaxation/viscosity as determinants of LV filling. The observed results provide preliminary data justifying further investigation of gender differences for parameters (stiffness vs. relaxation) and the extent to which they are altered in diabetic subjects. In addition, the conventional indexes of Epeak, Apeak, E/A, AT, DT, IVRT, E⬘, A⬘, E/E⬘, E⬘/A⬘, LVEDP, dP/dtmax, dP/dtmin and ␶ were determined for the diabetic and nondiabetic

Diastolic function in diabetes ● M. M. RIORDAN et al.

groups to assess their ability to differentiate between them. While this study is the first application of the PDF formalism to the diabetic human heart, Dent et al. (2001) applied the method to rats with streptozotocin-induced diabetes. Although the diabetic and control rats were not matched for HR, EF or MAP as were the groups in the present human study, important similarities were obtained. In exact concordance with this human study, Dent et al. (2001) obtained significantly higher values for c, kxo and 1/2kx2o in diabetic rats compared to nondiabetic controls, while no significant differences were observed for the other PDF parameters (␤ was not computed) (Dent et al. 2001). These E-wave based findings suggest that the same, or similar, pathophysiologic mechanisms are present affecting the chamber in both the animal model of diabetes and the presence of historically established diagnosis of diabetes in normotensive humans. Physiologically, the simultaneous process of active muscle relaxation and recoil generates the pressure gradient between the left atrium and LV that makes the LV initiate filling by mechanical suction. An alternative, fluid mechanics based rather than kinematics based characterization of the atrioventricular pressure gradient (kxo) is the Vp based computation of the intraventricular pressure gradient due to LV suction via the Bernoulli equation from color M-mode Doppler (Thomas et al. 1991). The PDF formalism models this process as the recoil of a damped SHO. Because the model is linear, the parameters c, k and xo can be determined uniquely (Kovács et al. 1987). Furthermore, additional indexes having practical physiological analogs can be derived. These include kxo (maximum force), 1/2kx2o (stored energy) and ␤ (c24mk) (relative influence of damping vs. stiffness). An important and practical consequence of the approach is that it allows determination of the extent to which relaxation (c) and stiffness (k) jointly contribute to E-wave DT (Kovács et al. 1997), rather than ascribing the entire DT purely to the effects of chamber stiffness (Little et al. 1995). The approach allows the determination of (mathematically) unique and (ideally) independent parameters that, while lumped, correspond to well-established aspects of the physiology of LV DF. The kinematic modeling approach for transmitral flow generates testable predictions. In kinematic terms, the product kxo denotes the initial peak force in the spring and is the analog of the maximum atrioventricular pressure gradient that generates the E-wave by mechanical suction. This prediction has been recently validated via simultaneous echo-cath data that shows that kxo is more consistent than 4V2 in predicting the instantaneous maximum pressure gradient (Bauman et al. 2004). Previous work has also shown k to reflect LV chamber stiffness (Kovács et al. 1997; Lisauskas et al. 2001), a property

1593

which is governed in part by the extracellular connective tissue matrix and by the intracellular protein titin (Granzier and Lebeit 2004). It is interesting to note that while no significant differences in xo and k individually were found between the controls and diabetics as a group, a significant increase in kxo was observed for the diabetic group. The ability to determine k and xo independently and then combine them into a physiologically meaningful term kxo helps to illustrate the manner in which diabetes may affect early LV filling: by requiring a greater pressure gradient to overcome increased viscoelastic effects without significantly altering chamber stiffness as assessed by k. It should also be noted that ␶ was not significantly different between groups, and, therefore, could not be responsible for the increased pressure gradient, as expressed by kxo, in the diabetic heart. In kinematic terms, 1/2kx2o represents the potential energy in the spring before release. Its physiological analog is the stored elastic strain energy available at mitral valve opening that generates chamber recoil and thereby generates the E-wave. The significantly greater values for kxo and 1/2kx2o in the diabetic group suggest an adaptive mechanism by the diabetic chamber to maintain a normal stroke volume and/or EF, but at the cost of increased energy utilization. It is consistent that a greater atrioventricular pressure gradient kxo (and consequently, 1/2kx2o) would be generated in response to increased damping (resistive losses) of transmitral flow, as manifested by a greater value of c. In kinematic terms, the parameter c represents the lumped viscoelastic (resistive) properties of the system. Its physiological analog includes any source of energy loss that opposes motion during filling. Several factors can influence filling through energy loss and therefore manifest as an increased value of c. These include the viscosity of the extracellular matrix, delayed relaxation, dynamic friction during sarcomere lengthening as the detached myosin heads slide past the thin filaments, pericardial effects and blood viscosity (Dent et al. 2001; Kass 2003). The diabetic group exhibited a significantly higher value for c than did the control group, and the level of significance attained was considerably higher than that of any of the other PDF parameters or terms. Specifically, 12 of the 15 diabetic subjects exhibited a higher value for c than the highest value of c exhibited by a member of the control group. A high value of c, while holding the parameters xo and k fixed, manifests in the E-wave shape as a shortened AT, lengthened DT and the presence of an inflection point on the deceleration portion of the contour. We conjecture that the increased values of c in the diabetic group primarily stem from two sources: delayed relaxation of the myocytes and ventricular remodeling. Delayed myocyte relaxation due to im-

1594

Ultrasound in Medicine and Biology

Volume 31, Number 12, 2005

Fig. 1. Representative transmitral Doppler data with model-predicted fits superimposed. Contours were normalized to peak E-wave velocity to facilitate comparison of shape and DT. (a) Control subject; parameters are xo ⫽ 8.9 cm, c ⫽ 13.7g/s, k ⫽ 157g/s2; MSE ⫽ 2 cm2/s2. (b) Diabetic subject; parameters are xo ⫽ 20.7 cm, c ⫽ 30.8g/s, k ⫽ 260g/s2; MSE ⫽ 7 cm2/s2. Conventional determination for DT by linear extension of the deceleration portion yields 190 ms for both subjects. (c) Alignment of model-predicted fits at peak velocity (control E-wave gray, diabetic E-wave black). Note that although DT by conventional methods is the same for both images, model-predicted best fits—for the entire contour—are achieved for higher values of c for the diabetic subject. This result shows the current gold standard of DF analysis (DT) is insufficient for detecting alterations to the myocardium and suggests that c is a more sensitive index of relaxation than DT.

pairment of calcium reuptake via the sarcoplasmic reticulum may cause delay and damping in relaxation. Also, disorganization of the extracellular matrix in diabetic hearts has been reported by previous studies (Spiro and Crowley 1993; Thompson 1994) and, along with myocardial fibrosis and AGEs, may promote damping through frictional losses (Kass 2003). With regard to conventional echocardiographic indexes, the results reported by Dent et al. (2001) in a diabetic rat model are similar to the results we obtained in humans. No significant differences were found between diabetics and controls for Epeak, Apeak or DT. However, AT was found to be shorter in the diabetic group in both studies. This result is predicted on kinematic grounds because the effect on SHO motion of an increase in c is to advance the E-wave peak closer to its leading edge (i.e., shorten AT). The only differences for echocardiographic indexes in this study vs. the study of Dent et al. (2001) were in the E/A ratio and IVRT. Dent et al. (2001) reported a lower E/A and higher IVRT in the diabetic animals. The differences in these parameters between groups in the present study did not achieve statistical significance. These findings, especially the observed nonsignificant differences in IVRT values among the two groups, may in part be attributable to the inclusion of individuals with CAD in both the diabetic and the control group. While the insonification direction (mitral leaflet tips) was not specifically selected for the purpose of optimally calculating IVRT from the Doppler images in this study, the hemodynamic data collected allowed us to determine ␶ to high precision. We can, therefore, assert with confidence that ␶ does not differentiate between groups. Since the observed IVRT values also do

not differentiate between groups, the lack of a significantly prolonged IVRT in diabetics in this study is corroborated. It should be noted that other studies have also found no significant difference in E/A (Grossman et al. 1991; Lee et al. 1997; Andren et al. 1998) and IVRT (Hsu et al. 1997; Andren et al. 1998) between diabetics and controls. An important result of the present study is that increased chamber viscosity, defined by increased “damping”, was observed in the diabetic group despite no significant increase in DT compared with the controls. At first, this finding may seem counterintuitive. The explanation resides in how the damping constant c and stiffness constant k individually affect DT. Specifically, an increase in c or a decrease in k individually or in tandem can increase DT. In the present study, a highly significant difference was found for c whereas no significant difference was found for k, indicating that diabetes increases the damping, or viscosity, of the chamber with little or no effect on its kinematic stiffness. Application of the PDF method to fully elucidate the presence and severity of DD in terms of chamber stiffness and viscosity in other selected pathophysiologic states is in progress. Our results underscore the importance of the proper interpretation of DT. Rather than being solely due to chamber stiffness as originally suggested (Little et al. 1995), DT is comprised of two components, one due to relaxation/viscosity (via c) and another due to stiffness (via k). Importantly the PDF method is unique in its ability to allow determination of both of these components by MBIP of E-waves, and it provides the algebraic relation (c2⫺4mk) by which they determine E-wave

Diastolic function in diabetes ● M. M. RIORDAN et al.

1595

Table 4. PDF Parameters and computed indexes — Comparison by presence of CAD within and across groups

Quantity c (g/s) k (g/s2) x0 (cm) kx0 (dynes) 1/2kx20 (ergs) ␤ (g2/s2) Quantity c (g/s) k (g/s2) x0 (cm) kx0 (dynes) 1/2kx20 (ergs) ␤ (g2/s2)

CAD

Non-CAD

Significance within group

Significance across groups

(n ⫽ 7) 17.9 ⫾ 4.7 172 ⫾ 32 11.3 ⫾ 1.5 1930 ⫾ 430 11000 ⫾ 3200 ⫺316 ⫾ 104 (n ⫽ 7) 24.7 ⫾ 4.1 203 ⫾ 49 12.3 ⫾ 4.7 2480 ⫾ 1100 17100 ⫾ 14000 ⫺169 ⫾ 176

(n ⫽ 8) 17.5 ⫾ 4.9 193 ⫾ 68 10.3 ⫾ 2.5 1900 ⫾ 540 9880 ⫾ 4130 ⫺395 ⫾ 203 (n ⫽ 8) 27.0 ⫾ 4.9 228 ⫾ 53 11.8 ⫾ 2.2 2680 ⫾ 700 16300 ⫾ 6900 ⫺149 ⫾ 356

Controls P ⫽ 0.87 P ⫽ 0.47 P ⫽ 0.38 P ⫽ 0.91 P ⫽ 0.57 P ⫽ 0.37 Diabetics P ⫽ 0.36 P ⫽ 0.36 P ⫽ 0.82 P ⫽ 0.67 P ⫽ 0.89 P ⫽ 0.90

CAD P ⫽ 0.01 P ⫽ 0.18 P ⫽ 0.61 P ⫽ 0.24 P ⫽ 0.29 P ⫽ 0.08 Non-CAD P ⫽ 0.002 P ⫽ 0.26 P ⫽ 0.21 P ⫽ 0.03 P ⫽ 0.04 P ⫽ 0.11

shape. Hence, despite the presence of DTs of similar duration, the damping constant c is significantly increased in diabetic hearts and can distinguish these hearts from nondiabetic hearts based on the E-wave contour (Fig. 1). Limitations A potential limitation of this study relates to the absence of indicators of the severity (glucose, HbA1c) and specific duration and method of treatment (oral antihyperglycemic vs. insulin) of diabetes in each patient. Higher values of c may be associated with poorer control of and/or longer duration of diabetes. However, the fact that the variance in c was comparable in the normal and diabetic subjects and negligible overlap of values for c between groups was obtained suggests that this limitation is relatively minor. Nevertheless, our results justify further studies to characterize the relation between PDF parameters and the severity and duration of diabetes. Another potential limitation concerns the fact that subjects with catheterization-proven CAD were included. The inclusion of subjects with CAD may affect diastolic filling through ongoing ischemia; however, none of the subjects had critical stenoses, active ischemia or wall motion abnormalities as evidenced by normal LVEF as determined by ventriculography. The fact that no differences were found in PDF, hemodynamic or conventional echocardiographic parameters between subjects with and without CAD, both in the normal and diabetic group, strongly suggests that the presence of anatomical CAD did not affect our conclusions. Furthermore, when patients with CAD were eliminated from both groups (the groups were still well-matched demographically), significant differences in the same PDF parameters and indexes were maintained (see Table 4). In addition, we note that patients with other risk factors were excluded to maintain the historically established diagnosis of diabe-

tes as the primary differentiating feature between the two groups. Finally, there is no simple, catheterization based analog for the viscosity/relaxation parameter c, as there is for stiffness (⌬P/⌬V). However, previous evidence that diabetes is associated with an increase in AGEs and other viscosity-enhancing changes (Spiro and Crowley 1993; Thompson 1994; Kass 2003) supports the view that chamber viscosity/relaxation is altered before significant, irreversible alterations to the myocardium ensue, leading to the onset of systolic dysfunction classically characterized as “diabetic cardiomyopathy”. CONCLUSIONS DF assessment by analysis of E-waves via the PDF method generated indexes of chamber viscosity/relaxation and stiffness that differentiated between diabetics and age-matched nondiabetic controls better than conventional E-wave based (E/A, Epeak, Apeak, DT, IVRT) or catheterization based (␶) indexes. The results, in concert with results of a similar study of an animal model of diabetes, reinforce the observation that in the diabetic heart, PDF determined chamber viscosity is increased while chamber stiffness is not. The PDF method permits determination of the independent viscosity/relaxation and stiffness contributions to DT, assessment of their relative magnitudes, which determine E-wave duration, and determination of the analogs of the peak atrioventricular pressure gradient and stored elastic strain energy before valve opening. This kinematic modeling of DF based approach of E-wave analysis allows the manifestations of diabetic cardiomyopathy to be quantified noninvasively and expressed in terms having physiological significance, suggesting an increase in chamber viscoelastic properties. Acknowledgments—The authors appreciate Peggy Brown’s help in echocardiographic data acquisition and Andrew Bowman’s helpful

1596

Ultrasound in Medicine and Biology

comments regarding manuscript preparation.This study was supported in part by the Heartland Affiliate of the American Heart Association (Dallas, TX, USA), the Whitaker Foundation (Roslyn, VA, USA), the National Institutes of Health (HL54179, HL04023 Bethesda, MD, USA), the Barnes-Jewish Hospital Foundation and the Alan A. and Edith L. Wolff Charitable Trust (St. Louis, MO, USA).

REFERENCES Andersen NH, Poulsen SH, Helleberg K, et al. Impact of essential hypertension and diabetes mellitus on left ventricular systolic and diastolic performance. Eur J Echocardiography 2003;4:306 –312. Andren B, Lind L, Hedenstierna G, Lithell H. Left ventricular diastolic function in a population sample of elderly men. Echocardiography 1998;15(5):433– 450. Appleton CP, Hatle LK, Popp RL. Relation of transmitral flow velocity patterns to left ventricular diastolic function: new insights from a combined hemodynamic and Doppler echocardiographic study. J Am Coll Cardiol 1988;12(2):426 – 440. Appleton CP, Firstenberg MS, Garcia MJ, Thomas JD. The echoDoppler evaluation of left ventricular diastolic function: A current perspective. Cardiology Clinics 2000;18(3):513–546. Aronson D, Rayfield EJ, Chesebro JH. Mechanisms determining course and outcome of diabetic patients who have had acute myocardial infarction. Ann Intern Med 1997;126:296 –306. Bauman L, Chung CS, Karamanoglu M, Kovc´s SJ. The peak atrioventricular pressure gradient to transmitral flow relation: kinematic model prediction with in-vivo validation. J Am Soc Echocardiogr 2004;17(8):839 – 844. Bland JM, Altman DG. Statistics notes: measurement error proportional to the mean. BMJ 1996;313:106. Cosson S, Kevorkian JP. Left ventricular diastolic dysfunction: an early sign of diabetic cardiomyopathy? Diabetes Metab 2003;29:455– 466. Courtois M, Kovács SJ, Ludbrook PA. Transmitral pressure-flow velocity relation. Importance of regional pressure gradients in the left ventricle during diastole. Circulation 1988;78:661– 671. Dent CL, Bowman AW, Scott MJ, et al. Echocardiographic characterization of fundamental mechanisms of abnormal diastolic filling in diabetic rats with a parameterized diastolic filling formalism. J Am Soc Echocardiogr 2001;14(12):1166 –1172. Eucker SA, Lisauskas JB, Singh J, Kovács SJ. Phase plane analysis of left ventricular hemodynamics. J Appl Physiol 2001;90:2238 –2244. Feigenbaum H. Echocardiography (5th ed.) Baltimore: Lea and Febiger, 1994:677. Granzier HL, Lebeit S. The giant protein titin: a major player in myocardial mechanics, signaling, and disease. Circ Res 2004;94(3): 284 –295. Grossmann G, Schmidt A, Hauner H, et al. Diastolic ventricular function in type 1 diabetic patients: a study using Doppler echocardiography. Diabet Med 1991;8(3):208 –212. Hall AF, Kovács SJ. Automated method for characterization of diastolic transmitral Doppler velocity contours: early rapid filling. Ultrasound Med Biol 1994;20:107–116. Hall AF, Bettlach JA, Nudelman SP, Kovács SJ. Manufacturer and machine setting-based variation of parameters resulting from model-based image processing of echocardiographic transmitral Doppler velocity profiles. Proceedings of the IEEE Ultrasonics Symposium, San Antonio, TX. 1996:1205–1210. Hsu KL, Chiang FT, Lo HM, et al. Cardiac contractility in noninsulindependent diabetes mellitus evaluated using the relation between end systolic wall stress and velocity of circumferential fiber shortening. Jpn Heart J 1997;38(4):463– 471. Kass DA. Getting better without AGE: New insights into the diabetic heart. Circ Res 2003;92:704 –706. Kovács SJ Jr., Barzilai B, Perez JE. Evaluation of diastolic function with Doppler echocardiography: the PDF formalism. Am J Physiol 1987;252:H178 –H187.

Volume 31, Number 12, 2005 Kovács SJ, Setser R, Hall AF. Left ventricular chamber stiffness from model-based image processing of transmitral Doppler E-waves. Coron Art Dis 1997;8:179 –187. Kovács SJ, Rosado J, Manson McGuire AL, Hall AF. Can transmitral Doppler E-waves differentiate hypertensive hearts from normal? Hypertension 1997;30:788 –795. Kovács SJ, Meisner JS, Yellin EL. “Modeling of diastole” in diastolic function and dysfunction. Cardiology Clinics 2000;18(3):459 – 490. Lee M, Gardin JM, Lynch JC, et al. Diabetes mellitus and echocardiographic left ventricular function in free-living elderly men and women: The Cardiovascular Health Study. Am Heart J 1997;133(1):36 – 43. Lipton P. Testing hypotheses: prediction and prejudice. Science 2005; 307(14):219 –221. Lisauskas J, Singh J, Courtois M, Kovács SJ. The relation of the peak Doppler E-wave to peak mitral annulus velocity ratio to diastolic function. Ultrasound Med Biol 2001;27(4):499 –507. Lisauskas JB, Singh J, Bowman AW, Kovács SJ. Chamber properties from transmitral flow: prediction of average and passive left ventricular diastolic stiffness. J Appl Physiol 2001;91:154 –162. Little WC, Ohno M, Kitzman DW, Thomas JD, Cheng C. Determination of left ventricular chamber stiffness from the time for deceleration of early left ventricular filling. Circulation 1995;92(7): 1933–1939. Lo SS, Leslie RD, Sutton MS. Effects of type 1 diabetes mellitus on cardiac function: A study of monozygotic twins. Br Heart J 1995; 73:450 – 455. Meyer TE, Karamanoglu M, Ehsani AA, Kovács SJ. Left ventricular chamber stiffness at rest as a determinant of exercise capacity in heart failure subjects with decreased ejection fraction. J Appl Physiol 2004;97(5):1667–1672. Naqvi TZ. Diastolic function assessment incorporating new techniques in Doppler echocardiography. Rev Cardiovasc Med 2003;4:81–99. Paillole C, Dahan M, Paycha F, et al. Prevalence and significance of left ventricular filling abnormalities determined by Doppler echocardiography in young type I (insulin-dependent) diabetic patients. Am J Cardiol 1989;64:1010 –1016. Press WH, Flannery BP, Teukolsky SA, Vetterling WT. Numerical recipies: The art of scientific computing. Cambridge: Cambridge University Press, 1986:523–528. Raev DC. Which LV function is impaired earlier in the evolution of diabetic cardiomyopathy? An echocardiographic study of young type I diabetic patients. Diabetes Care 1994;17:633– 639. Regan TJ, Weisse AB. Diabetic cardiomyopathy. J Am Coll Cardiol 1992;19:1165–1166. Rich MW, Stitziel NO, Kovács SJ. Prognostic value of diastolic filling parameters derived using a novel image processing technique in patients ⱖ70 years of age with congestive heart failure. Am J Cardiol 1999;84:82– 86. Rodrigues B, Cam MC, McNeill JH. Metabolic disturbances in diabetic cardiomyopathy. Mol Cell Biochem 1998;180(1–2):53–57. Shehadeh A, Regan TJ. Cardiac consequences of diabetes mellitus. Clin Cardiol 1995;18:301–305. Spiro MJ, Crowley TJ. Increased rat myocardial type VI collagen in diabetes mellitus and hypertension. Diabetologia 1993;36(2):93– 98. Thomas JD, Newell JB, Choong CY, Weyman AE. Physical and physiological determinants of transmitral velocity: numerical analysis. Am J Physiol Heart Circ Physiol 1991;260(5 Pt 2):H1718 – H1731. Thompson EW. Quantitative analysis of myocardial structure in insulin-dependent diabetes mellitus: effects of immediate and delayed insulin replacement. Proc Soc Exp Biol Med 1994;205(4):294 –305. Zile MR, Baicu CF, Gaasch WH. Diastolic heart failure—abnormalities in active relaxation and passive stiffness of the left ventricle. N Engl J Med 2004;350(19):1953–1959.