Quantification of the functional consequences of atrial fibrillation and surgical ablation on the left atrium using cardiac magnetic resonance imaging

Abstract

OBJECTIVES

The effect of atrial fibrillation (AF) on left atrial (LA) function has not been well defined and has been largely based on limited echocardiographic evaluation. This study examined the effect of AF and a subsequent Cox-Maze IV (CMIV) procedure on atrial function.

METHODS

Cardiac magnetic resonance imaging (cMRI) was performed in 20 healthy volunteers, 8 patients with paroxysmal atrial fibrillation (PAF) and 7 patients with persistent or long-standing persistent atrial fibrillation (LSP AF). Six of the PAF patients underwent surgical ablation with the CMIV procedure and 5 underwent both pre- and postoperative cMRIs. The persistent or LSP AF patients underwent only postoperative cMRIs because all scans were performed with patients in normal sinus rhythm. Volume–time curves throughout the cardiac cycle and regional wall shortening were evaluated using the cine images and compared across groups.

RESULTS

Compared with normal volunteers, patients with PAF had significantly decreased reservoir contribution to left ventricular (LV) filling (P = 0.0010), an increased conduit function contribution (P = 0.04) and preserved booster pump function (P = 0.14). Following the CMIV procedure, significant reductions were noted with respect to reservoir and booster pump function, with corresponding increases in conduit function. These differences were more drastic in patients with persistent/LSP AF. Regional wall motion was significantly reduced by PAF in all wall segments (P < 0.05), but was not further reduced by the CMIV. Despite changes in LA function, LV function was preserved following surgery.

CONCLUSIONS

PAF significantly altered LA function and has a detrimental effect on regional wall motion. Surgical intervention further altered LA function, but the reasons for this are likely multifactorial and not entirely related to the lesion set itself.

INTRODUCTION

Atrial fibrillation (AF) is known to interfere with the normal mechanical functioning of the atrium, and its effects on cardiac function may persist even when patients are in sinus rhythm. Studies have shown that dilatation of the atrium occurs early in AF and is related to cardiovascular morbidity and mortality [1]. Contractile and structural remodelling also occur in AF, leading to significant fibrosis, hypertrophy and myolysis; reduced left atrial (LA) contractility and impaired transport function [2]. Relatively few studies have formally examined the effect of AF on atrial function. Several reports have shown a decreased left atrial ejection fraction in patients with AF [3].

The Cox-Maze procedure was introduced to restore normal sinus rhythm in patients with AF and has become the surgical gold standard in its treatment. Previous studies examining the effect of the Cox-Maze procedure on atrial function were based on echocardiography and were limited to demonstrating atrial function as evidenced by the presence or absence of an A-wave on echocardiography [4, 5]. There have been no studies to assess the effects of the Cox-Maze IV (CMIV) procedure on either global or regional left atrial function in patients in a more comprehensive manner.

The evaluation of atrial function has attracted much less attention than that of ventricular function, and published studies on normal and pathological atrial function are limited. Atrial function is conceptually more complex and has been more challenging to quantify than ventricular function. It is now well established that the LA converts the continuous pulmonary venous return into intermittent, high inflow into the left ventricular (LV) during mitral valve opening with three distinct functional components: reservoir, conduit and booster pump function [6, 7]. During ventricular systole, the LA serves as a ‘reservoir’ and undergoes passive expansion that is driven by the apical descent of the mitral annulus, resulting in the generation of the pulmonary venous Doppler S-wave. Reservoir function is affected by LA compliance and relaxation, and it is an important determinant of cardiac output (CO) [8]. As the mitral valve opens following completion of ventricular ejection, the LA functions as a ‘conduit’ for the passage of blood that flows directly from the pulmonary veins into the LV. Although there is inflow from the pulmonary veins into the LA during early diastole (represented by the Doppler D-wave), LA volume is actually decreasing during this time as evidenced by the Doppler E-wave [9]. This occurs because LA reservoir and conduit volumes enter the LV simultaneously during early diastole. The rate of LV relaxation is an important determinant of LA conduit function because of the suction effect generated by the ventricle [10]. The last phase is characterized by active LA contraction and constitutes the ‘booster pump’ function of the LA, which serves as an additional determinant of ventricular end-diastolic volume. The relative importance of this active emptying (as opposed to passive emptying) increases with normal aging and in patients with diseased ventricles [11]. This active ejection, often known as the ‘atrial kick’, corresponds to the transmitral Doppler echocardiographic A-wave and is always accompanied by a slight amount of reverse flow into the pulmonary veins (represented by the pulmonary venous Doppler A-wave).

The purpose of this study was to (i) non-invasively perform a comprehensive quantification of global and regional LA function in normal individuals and patients with AF, (ii) determine the effect of AF on LA function and (iii) determine the effect of surgical intervention with a CMIV on LA function. To achieve this, we employed cardiac magnetic resonance imaging (cMRI) both to assess global atrial function and to characterize regional function of the LA by quantifying segmental wall motion using a novel approach. MRI is an excellent tool for studying cardiac physiology because of its non-invasive nature, its high-resolution image quality and its ability to record cardiac motion with distinct contrast between the myocardium and the blood. It has a major advantage over echocardiography in that its three-dimensional data format allows accurate measurement of cardiac chamber volumes in any plane through the heart during the entire cardiac cycle without requiring geometric assumptions and estimations.

METHODS

Patient populations and data collection

This study enrolled 20 healthy control subjects and 15 patients with AF. All patients with paroxysmal atrial fibrillation (PAF) were enrolled prospectively, and those with persistent or long-standing persistent (LSP) AF were enrolled either prospectively or retrospectively. Eight patients had paroxysmal AF and 7 patients had persistent or LSP AF. cMRI scans were obtained in all normal patients, preoperatively in all PAF patients and postoperatively in the 6 PAF patients who went on to receive a CMIV operation, which was performed as previously described [12]. Patients with persistent AF were only scanned postoperatively from a lone CMIV because, by definition, they were in AF preoperatively, precluding our ability to obtain a scan that would be comparable with the postoperative one obtained while in NSR. Significant structural or functional disease and any contraindication to MRI (e.g. non-compatible biometallic implants and claustrophobia) excluded patients from enrolment. This study was approved by the Washington University Institutional Review Board.

Imaging technique

Subjects were scanned using a clinical 1.5-T Cardiovascular MRI system (Acheiva 1.5 T, Release 2.5.3, Philips Healthcare Systems; Best, Netherlands). Anatomical and functional cine images of the heart were acquired using a designated cardiac phase array coil and a retrospectively gated, breath-held, balanced turbo field echo method with parallel imaging that was adopted from the approach previously described by Bowman and Kovacs [9]. Survey images and standard planes were obtained for the horizontal long-axis (HLA), short-axis and LV outflow tract (LVOT) views. High-resolution cine loops of the HLA and LVOT views were then obtained while subjects held their breath (average of 10–15 s breath-holds, depending on the heart rate). Retrospective gating, a technique that does not require a pause at end diastole for prospective R-wave detection, allowed imaging of the entire R–R interval of the cardiac cycle.

The HLA view was used to scan short-axis cine stack images perpendicular to the HLA axis. Approximately 20 short-axis stacks were generated 8 mm apart with zero gap, spanning from the LV apex through the superior-posterior wall of the LA. Short-axis cine image stacks were obtained during breath-holds, and cine loops were obtained for each slice for a full cardiac cycle. Each short-axis cine loop was divided into 30 cardiac phases. The repetition time, echo time and flip angles were 3.0 ms, 1.5 ms and 60°, respectively. In-plane resolution was 1.41 mm obtained with a field of view of 32 cm and a matrix size of 192 × 256 interpolated to 256 × 256. Total scanning time was ∼45 min. Heart rate was recorded during the entire examination and the R–R interval was synchronized accordingly during image acquisition. The scans were analysed using the ViewForum software (Philips Healthcare Systems), Merge eFilm (Merge Healthcare, Inc., Milwaukee, WI, USA) and Scion Image (Scion, Frederick, MD, USA).

Assessment of global left atrial function

Global LA functional analysis was performed by computing the volume changes in the LA and LV throughout the full cardiac cycle. Volume–time relationships of the LA and LV were derived using the short-axis stacks. The endocardial contours of the LA and LV in each short-axis image slice at each phase of the cardiac cycle were manually traced (Supplementary Fig. S1) to calculate the cross-sectional area in pixels, which was then converted to square centimetre. Using the slice thickness of 0.8 cm, the segmental volumes were calculated in millilitre for each short-axis stack. Simpson's rule was applied to derive the total LA and LV volumes at each phase of the cardiac cycle by summing the segmental volumes of the respective heart chambers using the following equation:

$$\mathrm{Chamber}\mathrm{volume}=\left(\sum _{i=1}^n{A}_i\right)d$$ (1)

where n represents the number of short-axis image slices spanning the LA or LV, A_i is the area of the endocardial contour from the image slice at the ith level and d is the slice thickness. This process of determining total volumes at each phase of the cardiac cycle was semi-automated for patients analysed using the ViewForum software. Papillary muscles and the pulmonary veins were excluded from the tracings. The LVOT and left atrial appendage (LAA) (in preoperative and normal patients) were identified and included in the volumetry. All volume calculations were performed by two authors (Jason O. Robertson and Rochus K. Voeller). User-dependent variability in this method has been previously studied using 10 randomly selected short-axis images of the LV and LA and has been found to be negligible [9].

A typical cMRI LA volume–time curve with the corresponding electrocardiogram (ECG) tracing of a normal subject is illustrated in Fig. 1. As the LV enters its systolic phase after the QRS complex, the LA begins to passively fill with pulmonary venous return from Point A, the minimal LA volume (LA_min) at ventricular end diastole. Approximately halfway through the cardiac cycle, the LA reaches its maximal volume at ventricular end systole (LA_max, Point B) and the LA begins to passively empty its contents into the LV via the mitral valve opening. Point C refers to the mid-diastolic relative minimal volume (LA_{rel min}) at the end of passive LA emptying. Point D is the relative maximal volume (LA_{rel max}) immediately prior to atrial systole. A normal LV volume–time curve is also plotted in Fig. 1, where Point E represents the LV end-diastolic volume (LV_EDV) and Point F represents the LV end-systolic volume (LV_ESV). LV stroke volume (LVSV) equals LV_EDV − LV_ESV.

Figure 1:

LA- and LV-volume vs time curves. From the LV curve, stroke volume (E–F) and LVEF can be calculated. See text for details on LA curve points A through D. A: LA minimal volume, LA_min; B: LA maximal volume, LA_max; C: LA relative minimal volume, LA_{rel min}; D: LA relative maximal volume, LA_{rel max}.

Using the parameters extrapolated from the LA and LV volume–time curves, the global function of the LA was calculated in terms of LA booster pump volume (LABPV), LA reservoir volume (LARV) and LA conduit volume (LACV) contribution to LV filling. LARV was defined as LA_max − LA_{rel min} (B–C). The LABPV was defined as LA_{rel max} − LA_min (D–A). LACV could not be determined from the LA volume–time curve alone, and had to be calculated using the following equation:

$$\text{LVSV}=\mathrm{LABPV}+\mathrm{LARV}+\mathrm{LACV}$$ (2)

The LVSV must equal total inflow from the LA, so the LACV may be derived following the calculation of the LVSV, LABPV and LARV from the LV and LA volume–time curves, respectively.

In order to further quantify left atrial reservoir, conduit and booster pump function, several additional measurements were also computed, as previously reported by Järvinen et al. [13] and Spencer et al. [11] using the following equations:

$$\text{Cyclic LA volume change }(\text{L}{\text{A}}_{\mathrm{CC}})=\text{L}{\text{A}}_{\max }-\text{L}{\text{A}}_{\min }$$ (3)

$$\begin{array}{rr}& \text{LA percent total emptying }(\text{L}{\text{A}}_{\mathrm{PTE}})\end{array}=(\text{L}{\text{A}}_{\max }-\text{L}{\text{A}}_{\min })\text{/L}{\text{A}}_{\max }\times \text{100}$$ (4)

$$\text{LA expansion index }(\text{L}{\text{A}}_{\mathrm{EI}})=(\text{L}{\text{A}}_{\max }-\text{L}{\text{A}}_{\min })\text{/L}{\text{A}}_{\min }\times \text{100}$$ (5)

$$\begin{array}{r}\text{LA passive emptying percentage of total emptying }(\text{L}{\text{A}}_{\mathrm{PE}})\\ =(\text{L}{\text{A}}_{\max }-\text{L}{{\text{A}}_{\mathrm{rel}}}_{\max })\text{/}(\text{L}{\text{A}}_{\max }-\text{L}{\text{A}}_{\min })\times \text{100}\end{array}$$ (6)

$$\text{LA passive emptying index }(\text{L}{\text{A}}_{\mathrm{PEI}})=(\text{L}{\text{A}}_{\max }-\text{L}{{\text{A}}_{\mathrm{rel}}}_{\max })\text{/L}{\text{A}}_{\max }\times \text{100}$$ (7)

$$\text{LA active emptying percentage of total emptying }(\text{L}{\text{A}}_{\mathrm{AE}})=(\text{L}{\text{A}}_{\mathrm{rel}\max }-\text{L}{\text{A}}_{\min })\text{/}(\text{L}{\text{A}}_{\max }-\text{L}{\text{A}}_{\min })\times \text{100}$$ (8)

$$\text{LA active emptying index }(\text{L}{\text{A}}_{\mathrm{AEI}})=(\text{L}{\text{A}}_{\mathrm{rel}\max }-\text{L}{\text{A}}_{\min })\text{/L}{\text{A}}_{\mathrm{rel}\max }\times \text{100}$$ (9)

$$\text{LA active ejection fraction }(\text{L}{\text{A}}_{\mathrm{EF}})=(\text{L}{\text{A}}_{\mathrm{rel}\max }-\text{L}{\text{A}}_{\min })\text{/L}{\text{A}}_{\max }\times \text{100}$$ (10)

Normalization for body surface area (BSA) was performed for LA_CC, but since the remaining indices of LA function are expressed as percentages, adjustment for BSA was typically not required.

Assessment of regional left atrial wall motion

Regional LA function was assessed in all subjects by visualizing segmental wall motion using cine MRIs. The LA wall was divided into four segments: anterior, posterior, medial and lateral. The percent shortening of each of the LA wall segments was estimated by calculating the difference in the distance between the superior aspect of the LA (defined as the fixed reference point) and the corresponding position on the mitral annulus during atrial end systole and end diastole (Fig. 2). As shown in Supplementary Fig. S2, the HLA cine MRIs were used to visualize and measure the medial and lateral LA wall segmental motion, and the LVOT view was used to measure the anterior and posterior LA wall.

Figure 2:

The regional assessment of LA function. (A) A schematic representation of the heart demonstrates the piston-like movement of the mitral valve plane during the cardiac cycle. For purposes of this analysis, the left atrium was divided into four segments: anterior, posterior, medial and lateral. The length of each segment during LA systole (S) and diastole (D) was measured to give fractional shortening during the cardiac cycle. The reference point is stationary throughout the cardiac cycle. (B) This axial cross-section through the heart demonstrates how the HLA and LVOT cine MRIs were used to visualize the four LA wall segments. MV: mitral valve annulus; TV: tricuspid valve annulus; AV: aortic valve; PV: pulmonic valve.

This method of quantifying LA regional function was based on the observation that the superior, mediastinal aspect of the LA adjacent to the right pulmonary veins remains stationary during the cardiac cycle. The simplified mechanical motion of the LA is that of a piston; the mitral annulus acts as a piston moving superiorly and inferiorly during the cardiac cycle. Under normal conditions, the LA wall glides smoothly against the pericardium during this piston-like motion of the mitral annulus (Fig. 2).

Statistical analysis

Continuous and categorical variables are expressed as mean ± SD and as number and percentage, respectively, unless otherwise specified. Comparisons were performed using a two-tailed, Student's t-test for normally distributed, continuous variables and Fisher's exact test for categorical variables. A Mann–Whitney U-test was used for non-parametric data. Matched data were analysed using a paired t-test following the demonstration of normality using the method of Kolmogorov and Smirnov. Multiple groups were compared using a one-way analysis of variance (ANOVA). A P-value of <0.05 was considered statistically significant. Statistical analyses were performed using SYSTAT 13 and Sigma Plot 12.3 (Systat Software, Inc., Richmond, CA, USA).

RESULTS

Baseline demographics

Baseline patient demographics are described in Table 1. All normal subjects were volunteers without any past medical or surgical history, and they were not taking any medications at the time of the study. The subjects had normal physical examinations, blood pressures and 12-lead ECGs immediately prior to undergoing MRI. While no echocardiographic data were available for this group, no normal subject had any significant valvular disease or wall motion abnormalities detected on cMRI, and left atrial sizes were within normal limits.

Table 1:

Baseline characteristics for patients undergoing the Cox-Maze IV operation

	No AF (%, n = 20)	PAF (%, n = 8)	Persistent/LSP AF (%, n = 7)	P-valuea	P-valueb
Demographics and history
Age (years)	30 ± 5	60 ± 13	57 ± 8	<0.0001	0.42
Body mass index	25 ± 3	28 ± 4	31 ± 4	0.071	0.23
Female	40 (8)	50 (4)	14 (1)	1.00	0.60
Diabetes	0 (0)	0 (0)	14 (1)	N/A	0.47
NYHA class III or IV	0 (0)	0 (0)	29 (2)	N/A	0.20
Hyperlipidaemia	0 (0)	13 (1)	71 (5)	0.29	0.041
Hypertension	0 (0)	50 (4)	57 (4)	0.0034	1.00
Prior stroke	0 (0)	13 (1)	0 (0)	0.29	1.00
Chronic lung disease (moderate/severe)	0 (0)	0 (0)	0 (0)	N/A	N/A
Renal failure	0 (0)	0 (0)	0 (0)	N/A	N/A
Peripheral vascular disease	0 (0)	13 (1)	0 (0)	0.29	1.00
AF and preoperative cardiac function
Duration of AF (years)	N/A	3.8 ± 3.3	4.5 ± 3.3	N/A	0.77
LSP AF	0 (0)	0 (0)	86 (6)	N/A	N/A
Prior failed catheter ablation	N/A	13 (1)	57 (4)	N/A	0.12
Prior myocardial infarction	0 (0)	25 (2)	0 (0)	0.074	0.47
Preoperative ejection fraction (%)	Not available	59 ± 8	56 ± 7	N/A	0.35
AV regurgitation (moderate/severe)	0 (0)c	0 (0)	0 (0)	N/A	N/A
MV regurgitation (moderate/severe)	0 (0)c	0 (0)	0 (0)	N/A	N/A
TV regurgitation (moderate/severe)	0 (0)c	13 (1)	0 (0)	0.29	1.00
Hypokinetic left ventricular wall motion	0 (0)c	13 (1)d	14 (1)e	0.29	1.00
Preoperative left atrial size (cm ± SD)	<4.2c	4.1 ± 0.9	4.8 ± 0.6	N/A	0.093
Operative characteristicsf
Lone Cox-Maze IV	N/A	67 (4)g	100 (7)	N/A	0.19
Sternotomy	N/A	67 (4)	57 (4)	N/A	1.00
Box lesion set	N/A	100 (6)	100 (7)	N/A	N/A
Reoperation	N/A	17 (1)h	0 (0)	N/A	0.46
CPB time (min)	N/A	128 ± 16	136 ± 50	N/A	0.78
Cross-clamp time (min)	N/A	41 ± 8	55 ± 29	N/A	0.57
Time from surgery to cMRI (years)	N/A	0.2 ± 0.2 median (range): 0.2 (0.1–0.5)	1.8 ± 2.1 median (range): 1.3 (0.2–6.2)	N/A	0.018
Arrhythmia follow-upf
1-year postoperative freedom from ATAs	N/A	100 (6)	100 (6)	N/A	N/A
1-year postoperative freedom from ATAs and AADs	N/A	83 (5)	100 (6)	N/A	1.00

^aComparing no AF to PAF.

^bComparing PAF with persistent.

^cBased on cMRI imaging, rather than echocardiography.

^dWall motion abnormalities identified by echocardiography correlated with cMRI, where the patient was noted to have posterior basal and apical septal/inferior dyskinesis on the preoperative scan and posterior basal hypokinesis on the postoperative scan. This fits with this patient's history of prior myocardial infarction.

^eHypokinetic left ventricular wall motion that was identified on preoperative echo was not seen on the postoperative cMRI.

^fFor only those patients who received a Cox-Maze IV operation (n = 6 for PAF patients).

^gOne patient had a papillary fibroelastoma removed from the pulmonic valve and the other had a ligation of a fistula between the left anterior descending and pulmonary artery.

^hPrevious coronary artery bypass graft.

AAD: antiarrhythmic drugs; ATA: atrial tachyarrhythmias; AF: atrial fibrillation; MV: mitral valve; NYHA: New York Heart Association; PAF: paroxysmal atrial fibrillation; LSP: long-standing persistent; TV: tricuspid valve.

Compared with normal subjects, patients with PAF were on average 30 years old and more frequently had hypertension (Table 1). Two of these patients had prior myocardial infarctions, and one individual had areas of hypokinetic wall motion in the left ventricle. Two of these patients had concomitant surgery: one had a removal of a papillary fibroelastoma from the pulmonic valve and the other had ligation of a fistula between the left anterior descending and pulmonary arteries. Patients in the persistent/LSP AF group primarily suffered from LSP AF (85.7%), and compared with PAF patients, they more frequently had hyperlipidaemia.

Global assessment of left atrial function

Cardiac MRI-derived indices of reservoir, conduit and booster pump function are compared between groups by ANOVA in Table 2. Compared with normal controls, the older PAF patients demonstrated increases in the relative contribution of active emptying (LA_AE: 30.2 ± 8.7% vs 57.5 ± 9.2%, P < 0.0001) and decreases in the relative contribution of passive emptying (LA_PE: 58.0 ± 9.2% vs 42.5 ± 9.2%, P = 0.0050) to ventricular filling. Booster pump function was otherwise preserved in preoperative PAF patients in normal sinus rhythm, but all indices of booster pump function were significantly reduced following surgery. The percentage of atrial volume emptying with each cycle (accounting for both reservoir and booster pump volumes, i.e. the LA_TPE) was diminished from normal in the preoperative PAF group (41.2 ± 6.9 vs 26.6 ± 4.9%, P < 0.0001) and further reduced in the postoperative PAF group (26.6 ± 4.9% vs 15.9 ± 5.6%, P < 0.0001). This is mirrored in both the preoperative and postoperative PAF groups, with significant reductions in the percentage change of LA volume during ventricular systole (LA_EI) and in the postoperative group with reductions in cyclic volume change (LA_CC). When a matched analysis is performed for the five PAF patients who had cMRIs both pre- and postoperatively from a CMIV, similar results were observed (Table 3).

Table 2:

Comparison of global left atrial function between groups by ANOVA

Variable	No AF (n = 20)	Preoperative PAF (n = 7)	Postoperative PAF (n = 6)	Postoperative persistent /LSP (n = 7)	P-valuea	P-valueb	P-valuec
Reservoir function
CC (ml)	47 ± 15	40 ± 6	22 ± 8	16 ± 7	0.020	0.0080	0.37
CC (ml/m2)	24 ± 6	20 ± 3	10 ± 3	7 ± 3	0.033	0.0010	0.24
TPE (%)	57 ± 6	41 ± 7	27 ± 5	16 ± 6	<0.0010	<0.001	0.0020
EI (%)	135 ± 37	72 ± 20	37 ± 9	19 ± 8	<0.0010	0.033	0.28
LARV (%)	39 ± 12	24 ± 3	16 ± 9	14 ± 7	0.0010	0.15	0.83
Conduit function
PE%	58 ± 9	43 ± 9	54 ± 15	61.7 ± 16.7	0.0050	0.093	0.23
PEI%	33 ± 7	17 ± 4	14 ± 5	10 ± 6	<0.0010	0.36	0.25
LACV (%)	40 ± 12	51 ± 9	74 ± 11	80 ± 8	0.037	0.0010	0.29
Booster pump function
AE (%)	30 ± 9	58 ± 9	45 ± 13	38 ± 17	<0.0010	0.045	0.31
AEI (%)	28 ± 8	29 ± 7	14 ± 5	6 ± 3	0.82	0.0010	0.056
LAEF (%)	24 ± 5	24 ± 6	12 ± 5	6 ± 3	0.94	<0.0010	0.016
LABPV (%)	22 ± 5	26 ± 9	11 ± 5	6 ± 3	0.14	<0.0010	0.11

^aComparing no AF with preoperative PAF.

^bComparing preoperative PAF with postoperative PAF.

^cComparing postoperative PAF with postoperative persistent/left atrial.

PAF: paroxysmal atrial fibrillation; LSP: long-standing persistent AF; CC: cyclic volume change; TPE: total percent emptying; EI: expansion index; LARV%: percent contribution of reservoir volume to left ventricular filling; PE: passive emptying percentage of total emptying; PEI: passive emptying index; LACV%: percent contribution of conduit volume to LV filling; AE: active emptying percentage of total emptying; AEI: active emptying index; LAEF: left atrial ejection fraction; LABPV%: percent contribution of booster pump volume to left ventricular filling.

Table 3:

Comparison of global left atrial function between matched preoperative PAF and postoperative PAF patients

	Preoperative PAF (n = 5)	Postoperative PAF (n = 5)	P-values
Reservoir function
CC (ml)	40 ± 5	22 ± 9	0.0050
CC (ml/m2)	19 ± 2	11 ± 3	0.0070
TPE (%)	42 ± 8	28 ± 4	0.011
EI (%)	74 ± 22	39 ± 8	0.016
LARV (%)	25 ± 1	16 ± 10	0.095
Conduit function
PE%	47 ± 7	56 ± 16	0.24
PEI%	19 ± 3	15 ± 4	0.30
LACV (%)	51 ± 10	75 ± 12	0.010
Booster pump function
AE (%)	53 ± 7	43 ± 13	0.098
AEI (%)	28 ± 8	14 ± 5	0.0010
LAEF (%)	23 ± 6	13 ± 5	0.0020
LABPV (%)	24 ± 9	10 ± 5	0.0080

The paired t-test assumes that differences are sampled from a Gaussian distribution. This assumption was tested using the method of Kolmogorov and Smirnov, and in each instance passed the normality test.

PAF: paroxysmal atrial fibrillation; CC: cyclic volume change; TPE: total percent emptying; EI: expansion index; LARV%: percent contribution of reservoir volume to left ventricular filling; PE: passive emptying percentage of total emptying; PEI: passive emptying index; LACV%: percent contribution of conduit volume to left ventricular filling; AE: active emptying percentage of total emptying; AEI: active emptying index; LAEF: left atrial ejection fraction; LABPV%: percent contribution of booster pump volume to left ventricular filling.

Further differences were recorded between postoperative PAF and postoperative persistent/LSP AF patients. By ANOVA, significant differences were observed for LA_TPE and LA_EF (Table 1). When these groups were compared without multiple comparisons, further differences were noted in the LA_EI (P = 0.0082) and in the percentage of atrial preload volume ejected into the LV (LA_AEI, P = 0.014).

The relative percent contributions of the booster pump, reservoir and conduit volumes to LV filling are presented in Fig. 3 and Table 2. The contribution of conduit volume progressively increases with PAF and again following surgery. This was mirrored by reductions in the contribution of reservoir volume in all groups, but this change was only statistically significant in comparison with normal controls. The contribution of the booster pump volume was significantly reduced with surgery, but not with PAF. No statistically significant differences were observed in postoperative PAF patients compared with postoperative persistent or LSP patients.

Figure 3:

The percentage of left atrial contribution to left ventricular filling. Compared with normal volunteers (n = 20), patients with PAF (n = 7) had a significantly decreased contribution of reservoir function to left ventricular filling and an increased conduit function contribution with preserved booster pump function. Patients undergoing the Cox-Maze IV (n = 6) had reduced booster pump function and further increases in conduit function. No differences were observed between postoperative PAF patients and postoperative persistent/LSP AF patients (n = 7) with respect to percent contributions to left ventricular filling.

Assessment of regional wall motion

Regional wall contraction for normal subjects was relatively uniform throughout the atrium. A significant reduction in segmental wall contraction was seen in all areas in PAF patients compared with normal controls (Fig. 4A). Anterior (26 ± 7% vs 17 ± 7%, P = 0.0091), posterior (27 ± 8% vs 19 ± 7%, P = 0.019), medial (30 ± 9% vs 17 ± 4%, P = 0.0012) and lateral (30 ± 10% vs 16 ± 5%, P = 0.0031) wall shortenings were similarly diminished.

Figure 4:

Regional wall motion. (A) Regional wall motion was significantly reduced by PAF. (B) The Cox-Maze IV procedure did not significantly affect regional shortening, where P > 0.05 for all areas examined.

Regional wall motion following the CMIV procedure trended towards further reductions in the anterior and posterior walls. However, a paired analysis did not demonstrate any significant changes postoperatively for the anterior (16 ± 7% vs 8 ± 3%, P = 0.1505), posterior (18 ± 7% vs 12 ± 3%, P = 0.22), medial (15 ± 5% vs 14 ± 6%, P = 0.92) or lateral (17 ± 5% vs 16 ± 7%, P = 0.77) wall segments (Fig. 4B).

Assessment of left ventricular function

A matched analysis of indices of LV function was performed for PAF patients that had both pre- and postoperative cMRI scans (Table 4). There were no significant differences in terms of left ventricular end-diastolic volume (LVEDV), left ventricular end-systolic volume (LVESV), left ventricular stroke volume (LVSV) or left ventricular ejection fraction (LVEF). However, CO was significantly increased following the CMIV procedure (P = 0.0090). This is primarily due to the fact that the heart rate increased following surgery (58 ± 15 vs 73 ± 10, P = 0.045).

Table 4:

Comparison of left ventricular function for matched preoperative PAF and postoperative PAF patients

	Preoperative PAF (n = 5)	Postoperative PAF (n = 5)	P-values
LVEDV (ml)	171 ± 49	176 ± 44	0.74
LVESV (ml)	77 ± 36	71 ± 28	0.39
CO (l/min)	5 ± 1	8 ± 1	0.009
LVSV (ml)	94 ± 25	104 ± 18	0.43
LVEF (%)	57 ± 12	61 ± 7	0.26

The paired t-test assumes that differences are sampled from a Gaussian distribution. This assumption was tested using a Shapiro–Wilkes test, and in each instance the variable passed the normality test.

PAF: paroxysmal atrial fibrillation; LVEDV: left ventricular end-diastolic volume; LVESV: left ventricular end-systolic volume; CO: cardiac output; LVSV: left ventricular stroke volume; LVEF: left ventricular ejection fraction.

DISCUSSION

The effects of AF and surgical correction of AF with the Cox-Maze procedure on left atrial reservoir, conduit and booster pump function are poorly studied, and previous reports are primarily based on limited two-dimensional (2D) echocardiographic analyses. The primary findings of this study are (i) paroxysmal AF results in preserved booster pump function and decreases in reservoir function that are offset by increases in conduit function, (ii) PAF is associated with significant decreases in regional wall motion, (iii) surgical correction of AF with a CMIV procedure is associated with further reductions in LA reservoir and booster pump function but no significant additional detriment in LA regional wall motion, (iv) postoperative persistent or LSP AF patients have more severe reductions in reservoir and booster pump function than postoperative PAF patients, and (v) postoperative changes in LA function do not negatively affect LV function in otherwise healthy individuals.

The differences in function that were found in this study between normal volunteers and PAF patients are likely due to a combination of age and PAF. The PAF cohort was significantly older, and it has been well established that age has a deleterious effect on atrial function. As patients age, they have been shown to become progressively more dependent on active, as opposed to passive, filling of the LV [11], which is consistent with the data we present that show an increase in LA_AE from normal patients to preoperative PAF patients. However, multiple echocardiographic studies have now demonstrated that conduit function deteriorates with age, while reservoir function is relatively preserved [11, 14]. This is in contrast to the cohort of PAF patients in this study who, despite being older, had a significantly increased conduit function in order to compensate for impaired reservoir function. Importantly, LA_EI and LA_PEI may be affected by increased minimum and total LA volumes, respectively, that could result from diastolic heart failure; however, there was no evidence of this in our patient cohorts. These data suggest an effect of PAF on atrial function that is independent of the changes seen with normal aging.

Furthermore, a striking difference between normal subjects and those with PAF became apparent with the analysis of regional wall motion. This analysis revealed a marked effect of AF on the segmental wall motion of the atria. Each segment of the atrial wall examined lost around 30–40% of its ability to shorten during the cardiac cycle when comparing the PAF patients with the normal controls. The fact that the percent active emptying appropriately increased and left atrial ejection fraction was preserved implies that the major change in atrial function is attributable to the decreased reservoir function and passive stretch, which are not typical of aging alone. This is likely because of the increased atrial fibrosis that occurs with structural remodelling from AF [2].

While this study also identified a significant detrimental effect of surgery on LA booster pump and reservoir function, it is unclear from these data whether this is a result of the lines of ablation, pericardiotomy, excision of the LAA or some combination thereof. Previously, the Cox-Maze procedure has been shown to have significant negative effects on LA function [4, 5]. The majority of these functional studies, however, were performed following a ‘cut-and-sew’ Cox-Maze III using 2D echocardiography, which provided limited functional data and no precise quantification of the reservoir, conduit and booster pump function of the LA. It has been estimated that isolation of the posterior left atrium in the original Cox-Maze III lesion set eliminated around 29% of the LA by weight and 35% of the LA by surface area from contributing to transport function [15]. While this may have an effect, based on the cMRI images, the posterior LA is not the only contributor to the impaired function that is seen following the CMIV: the changes are more global.

More recently, a porcine study that compared a modified Cox-Maze lesion set using bipolar radiofrequency (RF) ablation with pericardiotomy (a sham group) demonstrated that both groups had similar postoperative reductions in cMRI indices of booster pump and reservoir function, and there were no significant differences between the Cox-Maze and sham groups [16]. This suggests that sternotomy and pericardiotomy alone negatively influence normal LA function. This could be a result of the inflammation and scarring that occur following pericardiotomy. It is known that pericardiectomy results in increases in LA reservoir and booster pump function, among other macrophysiological effects, so it could be deduced that an intervention that causes more restriction on the movement of the heart or the coupling of the atria and ventricles would have the opposite effect [17].

Moreover, the LAA is routinely excised as part of the Cox-Maze lesion set in order to reduce risk of stroke, but several groups have advocated for the importance of the LAA to normal LA transport function. It has been demonstrated that the LAA accounts for as much as 17–30% of the total LA volume [18], and it has been reported that the LAA may be an important reservoir to protect against the increases in atrial pressure [18, 19]. The LAA is more compliant than the main left atrial chamber and might improve LA reservoir function under physiological conditions. For this reason, some have advocated a bilateral appendage-preserving maze procedure (BAP-Maze) [18]. In an echocardiographic study comparing patients with chronic AF who underwent a BAP-Maze (n = 46) to a similar cohort that underwent a Cox-Maze III, the BAP-Maze group had a significantly higher ratio of peak velocity of the A and E waves (0.52 ± 0.22 vs 0.25 ± 0.19, P < 0.0001). While this does not represent normalization of atrial transport function in either group, it clearly demonstrates improvement with the BAP-Maze. This group also demonstrated good LAA ejection fractions (44.7 ± 11.5%) following the procedure, which they contend may minimize the chance of thromboembolism from the LAA. The clinical benefit of this increase in atrial transport function remains unclear. As a result, we still advocate excision of the LAA due to its proven benefit in minimizing the long-term risk of stroke while off of anticoagulation [20].

With respect to regional wall motion changes following surgery, a matched analysis between pre- and postoperative PAF patients demonstrated no significant differences. It is possible that a difference would be observed with a larger sample size, but it does at least suggest that the effect of surgery on contractility is not as great as the effect of AF itself. This is consistent with a study that utilized multislice computed tomography to compare chronic AF patients undergoing mitral valve surgery and endocardial RF ablation with patients in sinus rhythm undergoing coronary artery bypass graft [21]. In that study, the extent of wall motion abnormalities in almost all segments of the LA wall were significantly worse in the AF group despite worse left ventricular function in the coronary artery bypass graft group, which would have affected LA function. The effect of the mitral valve operation/disease on this outcome versus the effect of AF and ablation is unclear, and no preoperative data are available for comparison. However, taken together with our data, which show a minimal effect of a more extensive lesion set on postoperative wall motion, these data suggest that the disease process is primarily responsible for the wall motion abnormalities.

While the function of the LA changes following a CMIV, it is important to understand whether those changes translate into reductions in LV function. Our data in a matched patient cohort suggest that LV function is preserved in postoperative PAF patients. In fact, CO increases following a CMIV due to the increases in basal heart rate, which may be due to a reduction in vagal input to the sinoatrial node after ablation. It is unclear whether these changes in LA function would affect LV function in patients with heart failure who rely more heavily on booster pump function.

Unfortunately, patients who were in persistent or LSP AF were, by definition, in AF prior to surgery, making meaningful pre- and postoperative comparisons of function impossible because measurements would be confounded by the arrhythmia. However, because this group was similar in most other respects to the postoperative PAF group, comparison between the two postoperative groups lends some insight into the effect of persistent/LSP AF on atrial function. This group did demonstrate further reductions in reservoir and booster pump function that would be consistent with the greater remodelling and fibrosis that are expected from more burdensome AF [2, 22]. This may be in part due to an increased number of prior catheter ablations in this group; however, previous reports have suggested beneficial remodelling of the left atrium without any detriments to function following catheter-based ablation [23, 24].

The ability to assess left atrial function, particularly pathophysiological changes, could provide important clinical information for both AF and other processes that impact the LA. These data inform surgeons of the impact of both AF and surgical ablation on left atrial function. A comprehensive system of quantifying atrial function could be used to evaluate disease progression or to study the impact of novel ablation modalities and surgical lesion sets used to treat AF. Moreover, the presence of severely impaired LA function may impact the clinical decision-making regarding discontinuation of anticoagulation in patients following surgical ablation.

Limitations

The disadvantages of cMRI include its high cost; long image-acquisition and data-processing times; contraindications for patients with non-compatible, implanted metallic devices and the load-dependent nature of the data obtained. Moreover, in calculating the LABPV, the assumption was made that the retrograde reflux of LA blood into the pulmonary veins during atrial systole (D–A, Fig. 1) was minimal, and so it was not factored into calculating the LABPV. This has been shown to be true in normal subjects [25]. In the diseased state, this assumption may not hold true and is a limitation of this study. Also, not all patients were scanned at the same interval following surgery. While we did not observe a time-dependent effect on postoperative LA function within groups, this non-standardized follow-up is a potential confounder of the comparison between the postoperative PAF and postoperative persistent/LSP cohorts. Finally, the normal patients in this study were much younger than the AF patients.

With respect to the regional wall motion analyses, it must be kept in mind that the accuracy of this method is dependent upon having normal LV structure and function, as mitral valve excursion can be altered by ventricular abnormalities independent of the LA. Patients with PAF were not analysed using this method because, while ejection fractions were normal, it is unlikely that after several years of persistent AF, those patients have ventricular structure and function that would be comparable to normal individuals or patients with PAF, and the absence of preoperative cMRI scans prevented us from using them as their own controls. For the matched analysis for pre- and postoperative PAF patients, MRI and echo data were evaluated to ensure that there were no changes in ventricular function following surgery. No such controls were possible with the comparison between PAF and normal patients, but MRI and echo data suggest that there were no major LV pathologies that would have confounded our results.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

Funding

This work was supported by National Institutes of Health grants 5R01 HL032257, R01 HL085113 and T32 HL07776.

Conflict of interest: Ralph J. Damiano and Richard B. Schuessler receive research grants and educational funding from AtriCure, Estech and Edwards. Ralph J. Damiano also receives consultant fees from AtriCure and Medtronic.