| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Krannert Institute of Cardiology, Indiana University School of Medicine, Indianapolis, Indiana 46202
Submitted 7 January 2004 ; accepted in final form 15 March 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
atrial conduction; optical mapping
5% of people over 65 yr old (5). Among the risk factors predisposing to AF are congestive heart failure (CHF), hypertension, and mitral valve disease, including both stenosis and insufficiency (reviewed in Ref. 5). In general, atrial dilatation is associated with an increased occurrence of AF (17, 28). Prevailing theory suggests that inducibility of sustained AF requires either a shortening of refractoriness, increased dispersion of refractoriness, or conduction slowing (1). Other factors that have been implicated in the stabilization of AF include increased conduction heterogeneity and anisotropy (23). The presence of one or more of these proarrhythmic factors would thus represent a "substrate for AF," contributing to increased AF stability. Various animal models of AF have been developed to study AF substrate. Parasympathetic stimulation (either with direct vagal nerve stimulation or circulating cholinergic agents) results in shortening of atrial refractoriness and sustained AF (10, 16). Prolonged rapid atrial pacing (RAP) leads to shortening of refractoriness and sustained AF (18, 30). In a dog model of CHF due to rapid ventricular pacing, no changes in refractoriness were found, but increased conduction heterogeneity was observed with epicardial electrode plaque mapping, and inducibility of AF was increased (14).
We have previously shown that in a canine model of chronic atrial dilatation due to mitral regurgitation (MR), left atrial (LA) myocardial structure is altered and inducibility of sustained AF is increased (29). However, unlike the CHF model, with the use of high-density (256 electrodes with interelectrode spacing of 4 mm) epicardial mapping there were no demonstrable electrophysiological abnormalities during normal pacing to explain the increased inducibility of AF: neither conduction slowing nor heterogeneous or shortened refractoriness was found (29). On the basis of these data, the substrate for AF may be different in the MR model. Theoretical data suggest that nonuniform anisotropy may be important in stabilizing reentrant circuits of small dimension, as in AF (26). In the present study, we have used high-resolution optical mapping to test the hypothesis that conduction abnormalities occur on a smaller scale and that nonuniform anisotropy is enhanced in this model of chronic atrial dilatation.
While RAP models may be representative of AF without underlying structural heart disease, AF in this MR model would be more representative of AF with underlying structural heart disease. The purpose of the present study is to compare atrial electrophysiology in canine RAP and MR models. We present evidence from high-resolution optical mapping recordings of atrial conduction that the substrate for AF in both models is distinctly different.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
Surgical procedure for the MR model. MR was produced as described previously (29). In short, a catheter with a stainless steel hook on the tip was advanced through a femoral artery line. The catheter was attached to a chorda of the mitral valve apparatus and pulled back to rupture the chordae, until MR was moderate to severe as judged by size and flow velocity of the regurgitant jet and acute LA dilatation was observed on transesophageal echocardiography. After creation of moderate to severe MR, animals were allowed to recover.
Surgical procedure for the RAP model. This procedure has been described in detail previously (21). In short, two pacing leads were inserted through the jugular vein and placed at the right ventricular apex and at the high right atrium (RA) at positions with a capture of <1 V. Complete atrioventricular (AV) block was created by radio-frequency ablation of the AV junction. The atrial lead was connected to a high-rate pulse generator (Itrel, Medtronic), and the ventricular lead was connected to a single-chamber ventricular pacemaker. The ventricular pacemaker was activated immediately at 100 beats/min, and 2 days after the surgical procedure, the atrial pacemaker was activated at a rate of 600 beats/min for chronic RAP.
Monitoring of the RAP and MR models. In the period before followup, all RAP and MR dogs were subjected to weekly physical exams. In addition, MR dogs were monitored by transthoracic echocardiography in the four-chamber view to measure LA length and ventricular function. MR dogs were followed up 37 ± 13 days after the creation of MR, at which time the LA length was 128 ± 9% of baseline, corresponding to an LA volume of 210 ± 47% of baseline. No physical signs of overt heart failure were observed in the study group.
In RAP dogs, the ECG was monitored weekly to ensure capture of the atrial and ventricular pacemakers. The atrial pacemaker was turned off temporarily to test for the presence of sustained AF. RAP dogs were followed up after 6–8 wk of rapid pacing.
Optical mapping.
After intravenous injection of 5,000 units of heparin, dogs were euthanized by injection of 50 mg/kg pentobarbital sodium, and hearts were rapidly excised and immersed in cardioplegic solution. The aorta was cannulated and perfused retrogradely with cardioplegic solution. The ventricles were then removed at 1 cm below the AV ring. After removal of the aortic cannula, the aortic wall was incised around the coronary ostia. Separate perfusion and pressure monitoring lines were inserted in both the right coronary and circumflex arteries and fixed in position by sutures around the coronary ostia. To ensure adequate atrial perfusion, all ventricular coronary branches were tied off.
Atrial preparations were transferred to a tissue chamber maintained at 37°C. The perfusion lines in both coronaries were perfused with normal solution (see Composition of salines). Pressure lines were connected to a pressure monitor, and flow through the perfusion lines was adjusted to maintain a pressure of 50–60 mmHg. Before optical recordings, normal saline containing 200 µg of the voltage-sensitive dye di-4-ANNEPS (Molecular Probes) was perfused through the preparation over a 15-min period.
Two-hundred fifty-six simultaneous optical action potentials were recorded with a 16 x 16 photodiode array from 2 x 2-cm areas on the anterior LA and RA appendage and free wall, using a setup described previously (31). The areas around the superior vena cava, posterior LA and RA, and Bachmann's bundle were excluded as recording locations. The approximate recording locations on the RA and LA are shown in Fig. 1. In total, LA locations were investigated in 9 control, 10 MR, and 7 RAP dogs. For the RA, seven locations were investigated in each group. During optical recordings, contractility was blocked with 15 mM butadione monoxime (BDM). After a recording location was selected, six unipolar stimulation electrodes were inserted around the field of view.
|
Composition of salines. Cardioplegic solution contained (in mmol/l) 123 NaCl, 15 KCl, 22 NaHCO3, 0.65 NaH2PO4, 0.50 MgCl2, 5.5 glucose, and 2 CaCl2, saturated with 95% O2-5% CO2. Normal solution contained (in mmol/l) 123 NaCl, 5.4 KCl, 22 NaHCO3, 0.65 NaH2PO4, 0.50 MgCl2, 5.5 glucose, and 2 CaCl2, saturated with 95% O2-5% CO2.
Data analysis. Optical signals were analyzed using custom-written software (M. Biermann). Local activation was determined as the time point of maximum change in fluorescence over time (dF/dt) for each fluorescent signal in the array. The baseline for each signal was set by interpolating between time points 25 ms before and 225 ms after each activation mark. The action potential duration (APD) at 80% repolarization (APD80) was determined as the time difference between the point of activation and the point during repolarization with 20% of the maximal action potential amplitude.
Conduction vector maps were constructed from the activation marks with custom-written software using an algorithm described earlier (4). Average conduction velocities (CVs) were calculated by taking the average of CVs in each element of the array determined from vector maps. APDs, CVs, and phase differences were calculated for one representative stimulus site at each recording location.
Phase difference (in ms/mm), an indicator of conduction heterogeneity (13, 14), was defined as the average difference in activation time between an element of the array with its neighboring elements divided by the distance between the elements. Frequency histograms were constructed for the phase differences within a recorded area. These histograms were summarized as the median phase time P50, and the 5th and 95th percentile, P5 and P95, of the distribution. The absolute degree of heterogeneity was quantified as the width of the distribution, P95 – P5. The relative heterogeneity was quantified as the width divided by the median, (P95 – P5)/P50.
All data are presented as means ± SD, unless mentioned otherwise. Statistical differences between the control, RAP, and MR groups in optical mapping experiments were tested by ANOVA analysis with a post hoc Student-Newman-Keuls test. A P value <0.05 was considered significant.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
|
|
|
|
|
|
In the MR LA, conduction was relatively homogeneous during normal pacing at a BCL of 350 ms (Fig. 5C, top row). By contrast, during extrastimulation at an S1–S2 of 160 ms (50 ms greater than the APD80), strong curvature of the activation wavefront became apparent, with regions of profoundly slow conduction. This was observed in all preparations in the MR LA, but not in the RA of the MR or either atria of the control or RAP preparations. The degree of conduction abnormality was dependent on propagation direction, with relatively normal conduction at some stimulation sites (Fig. 5C, last 2 columns) and highly disturbed conduction at others (Fig. 5C, first 2 columns), indicative of increased direction-dependence of conduction. The regions in the MR LA displaying slow conduction or strong wavefront curvature during extrastimulation did not have an abnormally long or short APD80 during normal pacing at 350 ms (Fig. 5C, bottom row).
To further quantify the relation between APD80 of the S1 and CVs during S1 and S2, linear regression analysis was performed for each recording location. Correlation of CVs with the APD80 of the S1 was very weak, both for the CV of the S1 [average correlation coefficients: control 0.052 ± 0.056; RAP 0.036 ± 0.050; MR 0.059 ± 0.052; P = not significant (NS)] and for the CV of the S2 (average correlation coefficients: control 0.050 ± 0.047; RAP 0.036 ± 0.050; MR 0.039 ± 0.047; P = NS, determined at an S1–S2 interval 50 ms above the APD80). For the RA, similar low values for the correlation coefficients were found.
Conduction in the RA did not show differences among the control, RAP, and MR groups (Fig. 6, A–C, respectively). Conduction during normal pacing at 350 ms was homogeneous and isotropic in all three models (Fig. 6, top row). APD80s were shorter in the RAP RA compared with the control and MR RA, without an increase in APD dispersion (Fig. 6, bottom row). Activation patterns during extrastimulation at an S1–S2 of 150 ms were similar to patterns during normal pacing (Fig. 6C, middle row). Thus, in contrast to extrastimulation in the MR LA, conduction of premature beats in the MR RA was comparable to conduction of premature beats in the control and RAP atria, indicating that conduction abnormalities during premature beats were confined to the LA in MR dogs.
|
Examples of phase difference maps during extrastimulation at a single coupling interval are shown in the middle row of Fig. 7 (summary data for all coupling intervals are shown in Fig. 8). During premature extrastimuli at most coupling intervals (e.g., 140 ms in Fig. 7), phase differences were low in the control LA (Fig. 7A, middle row). In RAP LA, phase differences during extrastimulation remained low, even at relatively short coupling intervals (e.g., 120 ms in Fig. 7B, middle row). By contrast, even at relatively long coupling intervals in the MR LA (e.g., 160 ms in Fig. 7C, middle row), regions with high phase differences became apparent, whereas other regions continued to have relatively low phase differences, indicative of increased conduction heterogeneity in the MR LA during extrastimulation. Interestingly, the degree of heterogeneity elicited by extrastimuli in the MR LA varied for different propagation directions.
For the control, RAP, and MR LA locations, respectively, the bottom rows of Fig. 7, A–C, illustrate frequency histograms for phase differences during the S2 depicted in the panels above (middle row). Whereas frequency histograms for the control and RAP are narrow, the histograms for S2s in the MR LA show a wider distribution, corresponding to increased conduction heterogeneity.
To quantify average conduction heterogeneity in all preparations, Fig. 8 shows plots of the parameters characterizing phase difference distributions for the LA (Fig. 8, left) and RA (Fig. 8, right) as a function of BCL during normal pacing. The median phase time P50 did not show significant differences between the groups in the LA (Fig. 8A, left). However, the absolute heterogeneity, P95 – P5, was significantly larger in the MR LA compared with the control LA (Fig. 8B, left) at short BCLs. Similarly, the relative heterogeneity, (P95 – P5)/P50, was significantly larger at short BCLs. The RAP LA did not significantly differ from the control LA in absolute and relative heterogeneity at any BCL. In the RA, neither the median phase time nor absolute and relative heterogeneity showed significant differences between the groups (Fig. 8, right).
In Fig. 9, the average parameters for phase distributions in all preparations during extrastimuli over a range of coupling intervals are plotted for the LA (Fig. 9, left) and RA (Fig. 9, right). Again, the RA did not show significant differences between the groups in median phase time and absolute or relative heterogeneity. By contrast, median phase time in the MR LA was significantly higher than in the control LA (Fig. 9A, left), corresponding to the regional conduction slowing apparent in activation maps. In addition, the absolute heterogeneity was strongly increased in the MR LA compared with the other groups, especially at shorter coupling intervals. The relative degree of heterogeneity was significantly higher in the MR LA compared with control at all coupling intervals lower than 200 ms. The RAP LA did not show significant differences from control at any coupling interval.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
Diverging structural as well as electrophysiological alterations have been reported for RAP (3, 32), CHF (14, 15), and MR models (29). Several studies have shown that in the RAP model, effective refractory periods (ERPs) (18, 30) and APDs (32) are shortened, without a change in CV. According to the multiple wavelet hypothesis (1), the decrease in ERP shortens the wavelength of reentrant wavelets, thereby allowing more wavelets to coexist in the atrial myocardium, leading to stabilization of AF. The RAP model is generally considered to be representative of AF without underlying structural heart disease in humans.
The canine model of CHF due to chronic rapid ventricular pacing and the MR model in the present study would be more representative for AF with underlying heart disease. In the CHF model, inducibility of sustained AF is increased (14). However, ERP and CV in open-chest studies on this model were not different from control. We have recently demonstrated that in our MR model, AF inducibility is also increased (29). Whereas CV was not significantly different in open-chest studies between control and MR atria, ERPs in the LA and RA of MR were prolonged compared with control. Thus, both in the CHF model and in the MR model, the increased inducibility of sustained AF cannot be explained by a decrease in wavelength.
In the CHF model, a marked increase in interstitial fibrosis has been observed in the LA and to a lesser extent in the RA (14). Epicardial mapping in open-chest studies revealed increased LA conduction heterogeneity. In the MR model, we have reported a more moderate increase in interstitial fibrosis in the MR LA, along with signs of chronic inflammation and fiber separation. However, during normal pacing in open-chest studies, epicardial mapping of conduction (with an interelectrode distance of 8 mm) did not indicate abnormal conduction in either atrium during normal pacing of the contralateral atrium (29).
In the present study, we have studied atrial conduction in more detail using high-resolution optical mapping during normal pacing and premature extrastimuli and have investigated whether there might be directional components to conduction abnormalities. As expected, APD80 in RAP atria was significantly reduced compared with control, corresponding to the electrical remodeling process that has been extensively characterized in earlier studies (reviewed in Ref. 8). By contrast, the LA and RA APD80 was not different between control and MR, indicating that the increase in ERP observed in open-chest studies was not caused by electrical remodeling at the cellular level. Instead, it is conceivable that the increased ERP in vivo is caused by an altered autonomic state secondary to the hemodynamic consequences of severe MR. The lack of effect of LA dilatation on the APD agrees with an earlier study on the lack of effect of naturally occurring LA enlargement on APD (9). However, the mitral valve stenosis in the group of 23 dogs included in that study was of unknown cause and duration and the study group had a heterogeneous background, which may have included CHF.
In control and RAP atria, conduction was homogeneous and relatively isotropic during normal pacing and extrastimulation. RAP atria did not show increased heterogeneity compared with control and displayed a similar degree of anisotropy. In fact, no differences were observed between control and RAP atria, except for the reduction in APD.
The MR RA was not significantly different from control in heterogeneity of conduction and anisotropy. By contrast, the MR LA showed increased conduction heterogeneity and anisotropy, especially during premature beats. Conduction abnormalities in the MR LA did not depend on differences in APD: dispersion of the APD in the MR LA was not higher than in the control LA, and the regions of slow conduction during premature beats were not associated with an abnormal APD during normal pacing. In principle, regional slow conduction during premature beats could result from regional variations in intrinsic membrane properties of myocytes, most importantly from regional reductions in sodium channel availability. However, such regional differences in intrinsic membrane properties would not be expected to display the dependence on propagation direction that was present in the MR LA. In addition, distributions of APD80 within recording locations were highly similar for different propagation directions during normal pacing, indicating that the differences observed for S2s originating from different stimulation sites are not caused by differences in the preceding S1s.
The divergence in MR LA and RA conduction paralleled structural abnormalities in the MR model. Whereas no structural differences from control were present in the MR RA, areas of increased interstitial fibrosis and fiber separation with signs of chronic inflammation were observed in the MR LA (29). We have confirmed that these structural alterations were indeed present within the optically mapped regions of the MR LA (not shown), suggesting that the regions of slow conduction correlate with regions of increased fibrosis or fiber separation. Because interstitial fibrosis and fiber separation would affect transverse propagation more than longitudinal propagation (23, 26), it would provide an explanation for the increased anisotropy of conduction.
This interpretation is compatible with theoretical and experimental studies by Spach and coworkers (23, 27), who have demonstrated that increased interstitial (micro)-fibrosis during aging and its concomitant decrease in side-to-side electrical coupling can cause a shift from uniform anisotropy to nonuniform anisotropy in atrial conduction. In nonuniform anisotropic tissue, very slow conduction may be observed during transverse propagation in the absence of variations in intrinsic membrane properties (25).
In areas of the atrial myocardium with highly organized underlying tissue structure like pectinate muscle or the crista terminalis, the shift from uniform to nonuniform anisotropy may cause slow transverse propagation even during normal pacing (24, 27). By contrast, although MR LA recording locations displayed increased anisotropy during normal pacing, very slow conduction was not observed during S1s and increased conduction heterogeneity was apparent only at shorter BCLs and during S2s.
Importantly, the structural alterations observed in the MR LA were regional: areas with increased interstitial fibrosis and fiber separation occurred alongside areas with relatively normal tissue organization (29). In this sense, structural alterations are more similar to the "patchy fibrosis" reported for cardiomyopathic human ventricle (12). The conduction patterns described in the present study for the MR LA are also reminiscent of those in cardiomyopathic human ventricle (12). There, in areas with patchy fibrosis, conduction was homogeneous during S1s, regardless of propagation direction. Conduction was homogeneous for S2s propagating in parallel to the main fiber orientation but could be strongly heterogeneous for S2s propagating perpendicularly.
The regions of increased conduction anisotropy and heterogeneity in the MR LA could provide a substrate for AF in the absence of APD shortening or increased APD dispersion. The description of atrial conduction in the MR LA presented in this study may be more representative than the RAP model for conduction in atria with underlying heart disease and conduction in aging atria (2, 11).
Limitations of this study.
Optical recordings revealed slowly conducting regions with reduced action potential amplitude and upstroke velocity of the optical action potential. It is important to note that because the optical action potential recorded for a single element of the array is the sum of optical action potentials in many myocytes, slow conduction in itself will cause a reduction in the amplitude and upstroke velocity of the ensemble optical action potential. To further elucidate the possible contribution of changes in cellular electrophysiology, intracellular recordings in perfused preparations and voltage-clamp measurements on isolated atrial myocytes are required, but this is beyond the scope of the present study.
The excitation-contraction uncoupler BDM employed in this study has been reported to shorten APD and reduce cycle length dependence of the APD (6, 31). However, the conduction abnormalities described in the present study were not related to the APD: areas in the MR LA of abnormal conduction during S2s did not have an abnormal APD during S1s. Moreover, these areas of abnormal conduction were already apparent at coupling intervals well above the APD80. Indeed, given the relative lack of cycle length dependence of the APD, the progressive increase in conduction heterogeneity with decreasing coupling intervals suggests that this heterogeneity was caused solely by underlying tissue substrate.
Finally, our recordings were made on 2 x 2-cm atrial areas on the anterior aspect of the appendages and free wall. We cannot exclude that other areas of the atrial myocardium, like Bachmann's bundle and the venous ostia, would be differentially affected and would have a divergent contribution to AF substrate formation in vivo.
In conclusion, high-resolution optical mapping shows that atria in the RAP model have relatively isotropic, homogeneous conduction comparable to the control atria. By contrast, chronic LA dilatation due to MR in the absence of heart failure caused increased conduction heterogeneity and anisotropy in the LA but not in the RA. Whereas the duration of the optical action potential was not affected in the MR model, it was reduced in the LA and RA of the RAP model. This study indicates that the underlying substrate for AF in chronically dilated atria is distinctly different from that of the RAP model. These differences in AF substrate could parallel divergent substrates for AF in humans and would indicate the desirability of differential treatment strategies.
|
GRANTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 Supplemental data for this article containing movie clips of LA conduction in the three groups may be found at http://ajpheart.physiology.org/cgi/content/full/00014.2004/DC1. 
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONGRANTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. C. Roberts-Thomson, I. H. Stevenson, P. M. Kistler, H. M. Haqqani, J. C. Goldblatt, P. Sanders, and J. M. Kalman Anatomically Determined Functional Conduction Delay in the Posterior Left Atrium: Relationship to Structural Heart Disease J. Am. Coll. Cardiol., February 26, 2008; 51(8): 856 - 862. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. M. Kistler, P. Sanders, M. Dodic, S. J. Spence, C. S. Samuel, C. Zhao, J. A. Charles, G. A. Edwards, and J. M. Kalman Atrial electrical and structural abnormalities in an ovine model of chronic blood pressure elevation after prenatal corticosteroid exposure: implications for development of atrial fibrillation Eur. Heart J., December 2, 2006; 27(24): 3045 - 3056. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. H. Everett IV, E. E. Wilson, S. Verheule, J. M. Guerra, S. Foreman, and J. E. Olgin Structural atrial remodeling alters the substrate and spatiotemporal organization of atrial fibrillation: a comparison in canine models of structural and electrical atrial remodeling Am J Physiol Heart Circ Physiol, December 1, 2006; 291(6): H2911 - H2923. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. W. Lee, T. H. Everett IV, D. Rahmutula, J. M. Guerra, E. Wilson, C. Ding, and J. E. Olgin Pirfenidone Prevents the Development of a Vulnerable Substrate for Atrial Fibrillation in a Canine Model of Heart Failure Circulation, October 17, 2006; 114(16): 1703 - 1712. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M. Guerra, T. H. Everett IV, K. W. Lee, E. Wilson, and J. E. Olgin Effects of the Gap Junction Modifier Rotigaptide (ZP123) on Atrial Conduction and Vulnerability to Atrial Fibrillation Circulation, July 11, 2006; 114(2): 110 - 118. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S.-i. Sakamoto, S. Yamauchi, H. Yamashita, H. Imura, Y. Maruyama, H. Ogasawara, N. Hatori, and K. Shimizu Intraoperative mapping of the right atrial free wall during sinus rhythm: variety of activation patterns and incidence of postoperative atrial fibrillation Eur. J. Cardiothorac. Surg., July 1, 2006; 30(1): 132 - 139. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. R. Ehrlich, S. H. Hohnloser, and S. Nattel Role of angiotensin system and effects of its inhibition in atrial fibrillation: clinical and experimental evidence Eur. Heart J., March 1, 2006; 27(5): 512 - 518. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H.-R. Neuberger, U. Schotten, Y. Blaauw, D. Vollmann, S. Eijsbouts, A. van Hunnik, and M. Allessie Chronic Atrial Dilation, Electrical Remodeling, and Atrial Fibrillation in the Goat J. Am. Coll. Cardiol., February 7, 2006; 47(3): 644 - 653. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||
