https://orcid.org/0000-0001-9511-8375
Abstract
Assessing left atrial (LA) size and function is an important part of the echocardiographic examination. We sought to assess how LA size and function develop over time, and which clinical characteristics promote atrial remodelling.
We examined longitudinal changes of the LA between two visits in the Copenhagen City Heart Study (n = 1065). The median time between the examinations was 10.4 years. LA measurements included: maximal LA volume (LAVmax), minimal LA volume (LAVmin), and LA emptying fraction (LAEF). Clinical and echocardiographic accelerators were determined from linear regression. The value of LA remodelling for predicting incident atrial fibrillation (AF) and heart failure (HF) was examined by Cox proportional hazards regressions. During follow-up, LAVmax and LAVmin significantly increased by 8.3 and 3.5 mL/m2, respectively. LAEF did not change. Age and AF were the most impactful clinical accelerators of LA remodelling with standardized beta-coefficients of 0.17 and 0.28 for changes in LAVmax, and 0.18 and 0.38 for changes in LAVmin, respectively. Left ventricular (LV) systolic function, diameter, and mass were also significant accelerators of LA remodelling. Changes in both LAVmax and LAVmin were significantly associated with incident AF [n = 46, ΔLAVmax: HR = 1.06 (1.03–1.09), P < 0.001 and ΔLAVmin: HR = 1.14 (1.10–1.18), P < 0.001, per 1 mL/m2 increase] and HF [n = 27, ΔLAVmax: HR = 1.08 (1.04–1.12), P < 0.001 and ΔLAVmin: HR = 1.13 (1.09–1.18), P < 0.001, per 1 mL/m2 increase].
Both maximal and minimal LA volume increase over time. Clinical accelerators included age and AF. LV structure and systolic function also accelerate LA remodelling. LA remodelling poses an increased risk of clinical outcomes.
Introduction
One of the key elements in the echocardiographic examination is the assessment of left atrial (LA) size, typically done through volume quantification.1 This provides information on the presence of diastolic dysfunction,2 serves as an indicator of the severity of mitral valve disease,3 and can be used to assess the risk of atrial fibrillation (AF) and heart failure (HF).4,5 Even though age-related alterations in diastolic flow measures are well-established, LA size has not been outlined as age-dependent in guidelines.2 Accordingly, changes in LA size may therefore be used as a marker of cardiovascular disease progression. However, the impression of no age-related change in atrial size is based on cross-sectional studies evaluating LA size by different age strata.6 There is a paucity of investigations examining the actual longitudinal changes through serial echocardiographic examinations. Furthermore, it is not established whether changes in LA size occur as part of the natural ageing process or only in the presence of cardiac risk factors. This is important to investigate in order to determine the impact of different cardiovascular risk factors and for guiding risk factor modification that could potentially reverse or halt LA remodelling.
Finally, there is accumulating evidence to suggest that measures of LA size and function other than the guideline-recommended measure, maximal LA volume (LAVmax), might serve as better tools for assessment of left ventricular (LV) filling pressure and to risk stratify patients.7,8 These alternative measures include the minimal LA volume (LAVmin) and LA emptying fraction (LAEF). Understanding how these measures develop longitudinally could give a better impression as to whether they may be more suitable than LAVmax for tracking pathological changes of the LA secondary to cardiovascular risk factors and LV abnormalities.
In light of the above, we sought to explore the following: (i) does remodelling of the LA occur as part of natural ageing, (ii) which clinical and echocardiographic risk factors accelerate changes of the LA, and (iii) are longitudinal changes in LA measures associated with an increased risk of incident AF and HF.
Methods
Population
We investigated participants included as part of a population-based study, the Copenhagen City Heart Study.9,10 We analysed data from participants who had a transthoracic echocardiogram with colour tissue Doppler imaging performed in the fourth study (n = 2221) who also underwent echocardiography in the fifth study (n = 1225). The fourth study was performed between 2001 and 2003 and the fifth study between 2011 and 2015. The fourth and fifth studies have been described previously.11 Participants who did not have LA measurements performed at baseline nor follow-up or who had missing details on body surface area were excluded (Figure 1). This left 1065 participants for final analyses.
Flow chart illustrating the study inclusion process for the overall purpose of investigating longitudinal changes in the general population, in the healthy subgroup (left panel), and for the prognostic assessment (right panel). AF, atrial fibrillation; CCHS, Copenhagen City Heart Study; HF, heart failure; IHD, ischaemic heart disease; IS, ischaemic stroke; LA, left atrium.
A subgroup analysis was performed to investigate longitudinal changes in healthy participants to investigate the impact of natural ageing. For this subgroup analysis, we further excluded participants with known cardiovascular disease or risk factors at the baseline visit, and also at the follow-up visit since these could have developed at any point in the interim between the examinations and thereby potentially have impacted the longitudinal changes (Figure 1). After all exclusions, 356 were eligible for this subgroup analysis.
Another subgroup analysis was performed to investigate the prognostic value of LA volume changes. In addition to the primary exclusions, we excluded participants with known AF or HF at baseline and follow-up (Figure 1). This left 979 for this analysis.
The study was performed in accordance with the second Helsinki Declaration, informed consent was obtained from all participants, and the study was approved by a regional scientific ethics committee.
Health examination
In addition to a transthoracic echocardiographic examination, the participants underwent a health examination at their study visits. The health examinations included a physical examination, a self-administered questionnaire, and blood samples. For this analysis, we considered the following biochemical parameters: cholesterol, HbA1c, haemoglobin, pro-brain natriuretic peptide (proBNP), and eGFR. Hypercholesterolemia was defined as either treatment with cholesterol-lowering medication or a total cholesterol > 6.5 mmol/L. Hypertension was defined as systolic blood pressure >140 mmHg, diastolic blood pressure >90 mmHg, or use of antihypertensive medication. Diabetes mellitus was defined as self-reported condition, plasma glucose ≥11.1 mmol/L, HbA1c >7%, or use of antidiabetic drugs. Details on smoking status and leisure physical activity were acquired from the self-administered questionnaire.
Echocardiography
The transthoracic echocardiograms were performed with GE Healthcare ultrasound machines (Vivid 5 for the fourth study and Vivid 9 for the fifth study). The examinations from the fourth study were stored on an external hard drive, whereas the examinations from the fifth study were stored in a remote digital image archive. Post-processing analyses were performed blinded to clinical data with commercially available software (EchoPAC 2008 and BT 113 for the fourth and fifth visits, respectively). The median time between the two examinations was 10.4 years (IQR: 10.2–10.8 years).
LV dimensions were measured by either M-mode or B-mode from the parasternal long-axis view at end-diastole. These measures included the interventricular septal diameter, LV internal diameter, and posterior wall diameter. These were used to calculate the LV mass by Devereux’ formula, which was indexed to body surface area to provide the LV mass index (LVMi).
Colour tissue Doppler loops were used to measure myocardial early relaxation velocity (e′) as an average of samples placed in all six LV walls. Pulsed-wave Doppler samples were placed at the mitral valve leaflets to obtain transmitral inflow profile. This was used to measure the peak early transmitral inflow (E-wave). The E-wave was indexed to e′ as a surrogate of LV filling pressure (E/e′).
LA volumes were measured by the biplane-area length method. The LAVmax was measured at end-systole, and the LAVmin at end-diastole. LAEF was calculated as: (LAVmax-LAVmin)/LAVmax. LA volumes in the fourth study were measured by a single investigator (TBS), and another investigator (NDJ) performed LA volume measurements in the fifth study.
LV speckle tracking was performed in the three apical projections. By placing three samples at the base of the LV and apex, a region of interest was generated from the software to track the LV myocardial speckles throughout the cardiac cycle. This generated a strain curve by which the peak global longitudinal strain (GLS) was derived. Speckle tracking was feasible in 694 (65%) of the participants at the baseline visit. The mean frame rate was 57 frames per second (minimum of 42 frames per second).
Outcome
The association between longitudinal changes in LA measures and relevant outcomes was investigated. The outcomes were incident AF and incident HF. Competing event was all-cause death. Endpoint data were extracted from the Danish National Patient Registry and the National Causes of Death Registry in July 2018.
Statistics
Changes in LA measures are reported as absolute differences and changes per year. The distributions of continuous measurements were assessed by QQ-plots. Due to skewness in LA measures, the Wilcoxon matched pairs signed-rank test was used to test for significant difference between follow-up and baseline measurements.
Linear regression analyses were applied to investigate potential clinical and paraclinical factors that could accelerate changes in LA volumes and to compare the impact using standardized beta coefficients. Potential echocardiographic measures at baseline that could accelerate LA remodelling were also examined in linear regressions. The candidate variables included GLS, LV diameter, LVMi, e′, and E/e′. Due to skewness, LVMi and E/e′ were log-transformed in these analyses. Overall, the linear regressions were performed in two steps. The base model included the respective baseline LA measure to account for regression to the mean. The second model included baseline LA measure and all clinical parameters: age, gender, diabetes mellitus, hypertension, hypercholesterolemia, ischaemic heart disease, HF, AF, smoking status, physical activity, heart rate, systolic and diastolic blood pressure, haemoglobin, and eGFR.
Using the same regression models, we also explored the impact of pro-brain natriuretic peptide (proBNP) on LA remodelling stratified by presence of AF due to effect-modification. Due to skewness, this variable was also log-transformed. Adjustments were made as in the models specified above.
Restricted cubic spline curves were constructed to visualize the continuous association between the dependent variables (ΔLAVmax and ΔLAVmin) and independent variables (age, GLS, LVIDd, and LVMi) in models adjusted for the variables in the second model. These spline curves allowed for investigation into potential non-linear associations. The number of knots was determined based on the lowest Akaike information criterion. In constructing these curves, the assumption of linearity was validated.
Similar unadjusted and adjusted linear regression analyses were performed to investigate the impact of age on change in LA measures (measured at baseline) from a cross-sectional perspective to examine whether such a cross-sectional regression would yield similar findings as for the longitudinal analyses.
Intra- and inter-observer analyses were performed in 20 randomly selected participants, and the results are reported as bias coefficients (mean difference ± SD).
Finally, we performed Cox proportional hazards analyses to investigate whether changes in LA measures were associated with the two endpoints of incident AF and incident HF. The time factor was from the follow-up visit (fifth study) until incident event or endpoint extraction in July 2018, whichever came first. Multivariable adjustments were made for clinical factors and relevant echocardiographic accelerators from the follow-up visit: baseline LA measure, age, gender, hypertension, ischaemic heart disease, hypercholesterolemia, BMI, smoking, haemoglobin, eGFR, LV internal diameter, LVMi, and GLS. Competing risk regression analyses were performed by the Fine and Grey method to account death as a potential competing event. Finally, restricted cubic spline curves based on Poisson regression were created to visualize the risk of AF and HF with increasing LA size. All statistical analyses were performed using STATA v. 15 SE (StataCorp LP, College Station, TX, USA). P-values <0.05 in two-tailed tests were considered significant.
Results
The clinical characteristics for the participants included in the fourth study and in the present analysis are shown in Table 1. Participants included in this analysis were younger and overall had fewer cardiovascular risk factors as compared with the inception cohort. For participants in this analysis, the mean age was 53 years, 45% were men, one-third had hypertension, and the mean blood pressure was 130/78 mmHg. During the median follow-up period of 10.4 years (IQR: 10.2–10.8 years), LAVmax significantly increased by 8.3 mL/m2, LAVmin significantly increased by 3.5 mL/m2, but LAEF did not change significantly. The annual increment was 0.79 mL/m2/year for LAVmax and 0.33 mL/m2/year for LAVmin. Even in the subset of healthy participants (n = 356), both LAVmax and LAVmin significantly increased although to a slightly lesser extent (LAVmax: 7.4 mL/m2, LAVmin: 3.1 mL/m2), with annual changes of 0.69 mL/m2/year and 0.30 mL/m2/year for LAVmax and LAVmin, respectively. Table 2 shows the detailed absolute and annualized longitudinal changes in LA measures.
Characteristics at the baseline visit
| All participants in CCHS4 echo visit | Participants included in this analysis | |
|---|---|---|
| n = 2221 | n = 1065 | |
| Clinical | ||
| Age (years) | 59 ± 16 | 53 ± 13 |
| Male gender, n (%) | 946 (43) | 481 (45) |
| Hypertension, n (%) | 984 (45) | 346 (33) |
| Diabetes mellitus, n (%) | 241 (11) | 79 (7) |
| Hypercholesterolemia, n (%) | 466 (24) | 215 (21) |
| Ischaemic heart disease, n (%) | 138 (6) | 42 (4) |
| Heart failure, n (%) | 30 (1) | 5 (1) |
| Atrial fibrillation, n (%) | 245 (11) | 62 (6) |
| Smoking, n (%) | ||
| Never | 695 (32) | 382 (37) |
| Previous | 744 (34) | 346 (33) |
| Current | 730 (34) | 314 (30) |
| Body mass index (kg/m2) | 26 ± 4 | 25 ± 4 |
| Heart rate (bpm) | 67 ± 12 | 66 ± 11 |
| Mean arterial pressure (mmHg) | 98 ± 14 | 95 ± 13 |
| Systolic blood pressure (mmHg) | 136 ± 23 | 130 ± 21 |
| Diastolic blood pressure (mmHg) | 78 ± 12 | 78 ± 12 |
| Biochemical | ||
| Haemoglobin (mmol/L) | 8.7 ± 0.7 | 8.7 ± 0.7 |
| HbA1c (mmol/mol) | 40 [37; 44] | 39 [36; 42] |
| Cholesterol (mmol/L) | 5.5 [4.7; 6.3] | 5.4 [4.7; 6.2] |
| eGFR (mL/min/1.73m2) | 83 [65; 104] | 92 [74; 110] |
| proBNP (pmol/L) | 17 [8; 31] | 14 [6; 24] |
| All participants in CCHS4 echo visit | Participants included in this analysis | |
|---|---|---|
| n = 2221 | n = 1065 | |
| Clinical | ||
| Age (years) | 59 ± 16 | 53 ± 13 |
| Male gender, n (%) | 946 (43) | 481 (45) |
| Hypertension, n (%) | 984 (45) | 346 (33) |
| Diabetes mellitus, n (%) | 241 (11) | 79 (7) |
| Hypercholesterolemia, n (%) | 466 (24) | 215 (21) |
| Ischaemic heart disease, n (%) | 138 (6) | 42 (4) |
| Heart failure, n (%) | 30 (1) | 5 (1) |
| Atrial fibrillation, n (%) | 245 (11) | 62 (6) |
| Smoking, n (%) | ||
| Never | 695 (32) | 382 (37) |
| Previous | 744 (34) | 346 (33) |
| Current | 730 (34) | 314 (30) |
| Body mass index (kg/m2) | 26 ± 4 | 25 ± 4 |
| Heart rate (bpm) | 67 ± 12 | 66 ± 11 |
| Mean arterial pressure (mmHg) | 98 ± 14 | 95 ± 13 |
| Systolic blood pressure (mmHg) | 136 ± 23 | 130 ± 21 |
| Diastolic blood pressure (mmHg) | 78 ± 12 | 78 ± 12 |
| Biochemical | ||
| Haemoglobin (mmol/L) | 8.7 ± 0.7 | 8.7 ± 0.7 |
| HbA1c (mmol/mol) | 40 [37; 44] | 39 [36; 42] |
| Cholesterol (mmol/L) | 5.5 [4.7; 6.3] | 5.4 [4.7; 6.2] |
| eGFR (mL/min/1.73m2) | 83 [65; 104] | 92 [74; 110] |
| proBNP (pmol/L) | 17 [8; 31] | 14 [6; 24] |
Continuous variables are expressed as mean ±SD for Gaussian distributed measures and median with inter-quartile ranges for skewed measures.
CCHS, Copenhagen City Heart Study; eGFR, estimated glomerular filtration rate; HbA1c, haemoglobin A1c.
Characteristics at the baseline visit
| All participants in CCHS4 echo visit | Participants included in this analysis | |
|---|---|---|
| n = 2221 | n = 1065 | |
| Clinical | ||
| Age (years) | 59 ± 16 | 53 ± 13 |
| Male gender, n (%) | 946 (43) | 481 (45) |
| Hypertension, n (%) | 984 (45) | 346 (33) |
| Diabetes mellitus, n (%) | 241 (11) | 79 (7) |
| Hypercholesterolemia, n (%) | 466 (24) | 215 (21) |
| Ischaemic heart disease, n (%) | 138 (6) | 42 (4) |
| Heart failure, n (%) | 30 (1) | 5 (1) |
| Atrial fibrillation, n (%) | 245 (11) | 62 (6) |
| Smoking, n (%) | ||
| Never | 695 (32) | 382 (37) |
| Previous | 744 (34) | 346 (33) |
| Current | 730 (34) | 314 (30) |
| Body mass index (kg/m2) | 26 ± 4 | 25 ± 4 |
| Heart rate (bpm) | 67 ± 12 | 66 ± 11 |
| Mean arterial pressure (mmHg) | 98 ± 14 | 95 ± 13 |
| Systolic blood pressure (mmHg) | 136 ± 23 | 130 ± 21 |
| Diastolic blood pressure (mmHg) | 78 ± 12 | 78 ± 12 |
| Biochemical | ||
| Haemoglobin (mmol/L) | 8.7 ± 0.7 | 8.7 ± 0.7 |
| HbA1c (mmol/mol) | 40 [37; 44] | 39 [36; 42] |
| Cholesterol (mmol/L) | 5.5 [4.7; 6.3] | 5.4 [4.7; 6.2] |
| eGFR (mL/min/1.73m2) | 83 [65; 104] | 92 [74; 110] |
| proBNP (pmol/L) | 17 [8; 31] | 14 [6; 24] |
| All participants in CCHS4 echo visit | Participants included in this analysis | |
|---|---|---|
| n = 2221 | n = 1065 | |
| Clinical | ||
| Age (years) | 59 ± 16 | 53 ± 13 |
| Male gender, n (%) | 946 (43) | 481 (45) |
| Hypertension, n (%) | 984 (45) | 346 (33) |
| Diabetes mellitus, n (%) | 241 (11) | 79 (7) |
| Hypercholesterolemia, n (%) | 466 (24) | 215 (21) |
| Ischaemic heart disease, n (%) | 138 (6) | 42 (4) |
| Heart failure, n (%) | 30 (1) | 5 (1) |
| Atrial fibrillation, n (%) | 245 (11) | 62 (6) |
| Smoking, n (%) | ||
| Never | 695 (32) | 382 (37) |
| Previous | 744 (34) | 346 (33) |
| Current | 730 (34) | 314 (30) |
| Body mass index (kg/m2) | 26 ± 4 | 25 ± 4 |
| Heart rate (bpm) | 67 ± 12 | 66 ± 11 |
| Mean arterial pressure (mmHg) | 98 ± 14 | 95 ± 13 |
| Systolic blood pressure (mmHg) | 136 ± 23 | 130 ± 21 |
| Diastolic blood pressure (mmHg) | 78 ± 12 | 78 ± 12 |
| Biochemical | ||
| Haemoglobin (mmol/L) | 8.7 ± 0.7 | 8.7 ± 0.7 |
| HbA1c (mmol/mol) | 40 [37; 44] | 39 [36; 42] |
| Cholesterol (mmol/L) | 5.5 [4.7; 6.3] | 5.4 [4.7; 6.2] |
| eGFR (mL/min/1.73m2) | 83 [65; 104] | 92 [74; 110] |
| proBNP (pmol/L) | 17 [8; 31] | 14 [6; 24] |
Continuous variables are expressed as mean ±SD for Gaussian distributed measures and median with inter-quartile ranges for skewed measures.
CCHS, Copenhagen City Heart Study; eGFR, estimated glomerular filtration rate; HbA1c, haemoglobin A1c.
Changes in left atrial measures
| Overall population | |||||
|---|---|---|---|---|---|
| n = 1065 | |||||
| Left atrial measures | Baseline visit | Follow-up visit | Absolute change | Annualized change | P-value |
| LAVmax (mL/m2) | 17.8 (14.5–21.7) | 26.5 (21.5–32.2) | 8.3 (3.9–13.8) | 0.79 (0.36–1.33) | <0.001 |
| LAVmin (mL/m2) | 8.2 (6.3–10.5) | 11.7 (9.2–15.1) | 3.5 (1.1–6.6) | 0.33 (0.10–0.63) | <0.001 |
| LAEF (%) | 54 (44–62) | 55 (48–61) | 0.25 (−8.73–12.00) | 0.02 (−0.81–1.12) | 0.76 |
| Healthy population | |||||
| n = 356 | |||||
| LAVmax (mL/m2) | 17.5 (14.3–21.0) | 25.0 (20.7–30.0) | 7.4 (3.3–12.4) | 0.69 (0.32–1.20) | <0.001 |
| LAVmin (mL/m2) | 7.7 (6.0–9.8) | 11.0 (8.9–13.5) | 3.1 (1.0–5.7) | 0.30 (0.10–0.55) | <0.001 |
| LAEF (%) | 55 (46–64) | 56 (50–62) | −0.03 (−8.60–11.31) | 0.00 (−0.83–1.03) | 0.87 |
| Overall population | |||||
|---|---|---|---|---|---|
| n = 1065 | |||||
| Left atrial measures | Baseline visit | Follow-up visit | Absolute change | Annualized change | P-value |
| LAVmax (mL/m2) | 17.8 (14.5–21.7) | 26.5 (21.5–32.2) | 8.3 (3.9–13.8) | 0.79 (0.36–1.33) | <0.001 |
| LAVmin (mL/m2) | 8.2 (6.3–10.5) | 11.7 (9.2–15.1) | 3.5 (1.1–6.6) | 0.33 (0.10–0.63) | <0.001 |
| LAEF (%) | 54 (44–62) | 55 (48–61) | 0.25 (−8.73–12.00) | 0.02 (−0.81–1.12) | 0.76 |
| Healthy population | |||||
| n = 356 | |||||
| LAVmax (mL/m2) | 17.5 (14.3–21.0) | 25.0 (20.7–30.0) | 7.4 (3.3–12.4) | 0.69 (0.32–1.20) | <0.001 |
| LAVmin (mL/m2) | 7.7 (6.0–9.8) | 11.0 (8.9–13.5) | 3.1 (1.0–5.7) | 0.30 (0.10–0.55) | <0.001 |
| LAEF (%) | 55 (46–64) | 56 (50–62) | −0.03 (−8.60–11.31) | 0.00 (−0.83–1.03) | 0.87 |
All variables are presented as the median value with inter-quartile range.
LAEF, left atrium emptying fraction; LAVmax, maximal left atrial volume; LAVmin, minimal left atrial volume.
Changes in left atrial measures
| Overall population | |||||
|---|---|---|---|---|---|
| n = 1065 | |||||
| Left atrial measures | Baseline visit | Follow-up visit | Absolute change | Annualized change | P-value |
| LAVmax (mL/m2) | 17.8 (14.5–21.7) | 26.5 (21.5–32.2) | 8.3 (3.9–13.8) | 0.79 (0.36–1.33) | <0.001 |
| LAVmin (mL/m2) | 8.2 (6.3–10.5) | 11.7 (9.2–15.1) | 3.5 (1.1–6.6) | 0.33 (0.10–0.63) | <0.001 |
| LAEF (%) | 54 (44–62) | 55 (48–61) | 0.25 (−8.73–12.00) | 0.02 (−0.81–1.12) | 0.76 |
| Healthy population | |||||
| n = 356 | |||||
| LAVmax (mL/m2) | 17.5 (14.3–21.0) | 25.0 (20.7–30.0) | 7.4 (3.3–12.4) | 0.69 (0.32–1.20) | <0.001 |
| LAVmin (mL/m2) | 7.7 (6.0–9.8) | 11.0 (8.9–13.5) | 3.1 (1.0–5.7) | 0.30 (0.10–0.55) | <0.001 |
| LAEF (%) | 55 (46–64) | 56 (50–62) | −0.03 (−8.60–11.31) | 0.00 (−0.83–1.03) | 0.87 |
| Overall population | |||||
|---|---|---|---|---|---|
| n = 1065 | |||||
| Left atrial measures | Baseline visit | Follow-up visit | Absolute change | Annualized change | P-value |
| LAVmax (mL/m2) | 17.8 (14.5–21.7) | 26.5 (21.5–32.2) | 8.3 (3.9–13.8) | 0.79 (0.36–1.33) | <0.001 |
| LAVmin (mL/m2) | 8.2 (6.3–10.5) | 11.7 (9.2–15.1) | 3.5 (1.1–6.6) | 0.33 (0.10–0.63) | <0.001 |
| LAEF (%) | 54 (44–62) | 55 (48–61) | 0.25 (−8.73–12.00) | 0.02 (−0.81–1.12) | 0.76 |
| Healthy population | |||||
| n = 356 | |||||
| LAVmax (mL/m2) | 17.5 (14.3–21.0) | 25.0 (20.7–30.0) | 7.4 (3.3–12.4) | 0.69 (0.32–1.20) | <0.001 |
| LAVmin (mL/m2) | 7.7 (6.0–9.8) | 11.0 (8.9–13.5) | 3.1 (1.0–5.7) | 0.30 (0.10–0.55) | <0.001 |
| LAEF (%) | 55 (46–64) | 56 (50–62) | −0.03 (−8.60–11.31) | 0.00 (−0.83–1.03) | 0.87 |
All variables are presented as the median value with inter-quartile range.
LAEF, left atrium emptying fraction; LAVmax, maximal left atrial volume; LAVmin, minimal left atrial volume.
Based on cross-sectional unadjusted linear regression, examining the changes in baseline LA measures with age, we observed that LAVmax increased with 0.35 mL/m2 per 10-year increase (P = 0.011), LAVmin increased with 0.62 mL/m2 per 10-year increase, and LAEF decreased with 1.96% per 10-year increase. After multivariable adjustments, only LAVmin and LAEF changed with increasing age [LAVmin: 0.43 mL/m2 (0.17, 0.68 mL/m2), P = 0.001; LAEF: −1.93% (−2.79, −1.06%), P < 0.001].
Accelerators of LA remodelling
In linear regression, several clinical risk factors accelerated LA remodelling, including age, AF, hypercholesterolemia, ischaemic heart disease, and systolic blood pressure, with age and AF showing the highest standardized beta coefficients (Figure 2A and B). Better kidney function by eGFR attenuated LA remodelling. Of note, hypercholesterolemia selectively accelerated changes in LAVmax. However, after adjusting for all clinical variables, only age and AF remained significant accelerators of LA remodelling for both LA measures, and hypercholesterolemia also remained a significant accelerator of remodelling of LAVmax (Figure 2C and D). The continuous association between changes in LA measures and age is shown in Figure 3A and B.
Impact of clinical characteristics and echocardiographic factors on change in LA volumes. The figures are forest plots showing the impact of clinical and echocardiographic variables on changes in LA volumes by standardized beta coefficients. (A) The estimated standardized beta coefficient of each clinical variable when adjusted for baseline LAVmax and (B) similarly shows the estimated standardized beta coefficient of each variable on change on LAVmin after adjusting for baseline LAVmin. (C and D) The fully adjusted estimated standardized beta coefficient after including all clinical variables and baseline LAVmax and LAVmin in (C and D), respectively. (E) The impact of key echocardiographic measures on change in LAVmax when adjusted for baseline LAVmax (model 1) and in a fully adjusted model including all clinical variables (model 2). (F) The impact of key echocardiographic measures on LAVmin in similar adjusted models. LAVmax, maximal left atrial volume; LAVmin, minimal left atrial volume.
Continued.
Continued.
Impact of age on change in LA volumes. The figures show the continuous association between age and change in LAVmax (ΔLAVmax) and change in LAVmin (ΔLAVmin) for the overall population. (A and B), respectively, show that LAVmax and LAVmin changes linearly with increasing age. The associations are adjusted for all relevant confounders (model 2). LA, left atrium; LAVmax, maximal left atrial volume; LAVmin, minimal left atrial volume.
Of note, proBNP was associated with an accelerated remodelling of LAVmin but not LAVmax. However, this association was significantly modified by presence of AF (P for interaction = 0.001), such that in participants without AF, increasing proBNP was associated with an accelerated remodelling of LAVmin, whereas in participants with AF it was associated with a decelerated remodelling of LAVmin. These findings were consistent after multivariable analyses in fully adjusted models [ΔLAVmin in those without AF: standardized β = 0.06 (0.00, 0.13), P = 0.048; ΔLAVmin in those with AF: standardized β=−0.34 (−0.67, −0.02), P = 0.040].
All investigated echocardiographic measures were significant accelerators of changes in LA measures when adjusted for baseline LA measure, although their impact on LAVmax and LAVmin was different (Figure 2E and F). However, in fully adjusted models, only GLS, LV size, and LVMi were significantly associated with changes in LA measures such that LV size was associated with changes in both LA measures (Figure 4A and B), whereas GLS was only associated with changes in LAVmax (Figure 4C), and LVMi was only associated with changes in LAVmin (Figure 4D).
Impact of echocardiographic factors on LA volumes. The figures show the continuous association between echocardiographic measures and LA volumes for the overall population. Specifically, they show left ventricular size (left ventricular internal diameter) and changes in LAVmax (A) and LAVmin (B), between global longitudinal strain and changes in LAVmax (C), and between left ventricular mass index and changes in LAVmin (D). All figures illustrate a linear acceleration in LA volume changes with increasing left ventricular size and mass and with deteriorating global longitudinal strain. The associations are adjusted for the respective baseline LA measure, age, and gender. LA, left atrium; LAVmax, maximal left atrial volume; LAVmin, minimal left atrial volume
Reproducibility
Inter-observer analyses revealed bias coefficients for LAVmax and LAVmin of 1.59 ± 3.68 and −0.23 ± 2.14 mL, respectively. The variability was lower in intra-observer analyses with bias coefficients for LAVmax and LAVmin of −0.65 ± 2.70 and −0.12 ± 1.47 mL, respectively.
Prognostic value of longitudinal changes in LA measures
Of the 979 participants in the prognostic analysis, 46 developed AF and 27 developed HF during a median follow-up time of 6.3 years (IQR: 5.3; 6.7 years). When adjusted for baseline LA measures, both ΔLAVmax and ΔLAVmin were significantly associated with incident AF [ΔLAVmax: HR = 1.06 (1.03–1.09), P < 0.001, per 1 mL/m2 increase, ΔLAVmin: HR = 1.14 (1.10–1.18), P < 0.001, per 1 mL/m2 increase]. Even after adjusting for clinical and echocardiographic confounders at the follow-up visit, changes in both LA measures remained significantly associated with AF [ΔLAVmax: HR = 1.07 (1.03–1.12), P = 0.001, per 1 mL/m2 increase, ΔLAVmin: HR = 1.13 (1.07–1.19), P < 0.001, per 1 mL/m2 increase, Figure 5A and B]. The associations were unchanged in competing risk regression when performing the same multivariable adjustments while accounting for all-cause death as a competing event (n = 61).
Risk of outcome according to changes in LA volumes. The figures show the risk of incident atrial fibrillation according to longitudinal changes in LAVmax and LAVmin (A and B, respectively) as well as the risk of incident heart failure (C and D). The associations are adjusted for baseline LA measure, age, gender, hypertension, ischaemic heart disease, hypercholesterolemia, body mass index, smoking, haemoglobin, eGFR, left ventricular internal diameter, left ventricular mass index, and global longitudinal strain. eGFR, estimated glomerular filtration rate; LA, left atrium; LAVmax, maximal left atrial volume; LAVmin, minimal left atrial volume.
When adjusted for baseline LA measures, both ΔLAVmax and ΔLAVmin were also significantly associated with incident HF [ΔLAVmax: HR = 1.08 (1.04–1.12), P < 0.001, per 1 mL/m2 increase, ΔLAVmin: HR = 1.13 (1.09–1.18), P < 0.001, per 1 mL/m2 increase]. These findings remained unchanged after adjusting for clinical and echocardiographic confounders [ΔLAVmax: HR = 1.07 (1.02–1.12), P = 0.002, per 1 mL/m2 increase, ΔLAVmin: HR = 1.10 (1.04–1.16), P = 0.001, per 1 mL/m2 increase, Figure 5C and D]. The associations remained unchanged after similar multivariable adjustments while accounting for all-cause death as competing event.
Discussion
This is the first echocardiographic study to examine long-term longitudinal changes in LA volumes. The principal findings from the present report were that (i) LA volumes increased significantly in individuals with risk factors as well as with natural ageing over time, (ii) age and AF are the most influential clinical accelerators of LA remodelling, and (iii) changes in LA volumes are independently associated with increased risks of both AF and HF.
Interestingly, we observed a marked discrepancy when examining the change in LA measures with age based on cross-sectional analyses as compared with the longitudinal analyses. The cross-sectional analyses indicated that LAVmin and LAEF changed with age, which is in contrast to the longitudinal analyses which found that only LAVmax and LAVmin changed during follow-up, and that these changes were substantially larger than the estimates derived from the cross-sectional results. This highlights the importance of performing such explorations in a longitudinal design.
In recent years, novel measures of LA function, specifically LA speckle tracking, have come into focus and shown clinical potential in terms of assessing diastolic dysfunction and cardiovascular risk.12,13 Similar longitudinal studies are needed in the future to explore how LA strain measurements develop over time and how such changes relate to prognosis.
Longitudinal changes in LA
Current guidelines for assessing diastolic function recommend age-appropriate cut-offs when considering diastolic measures.2 However, LA volume is not denoted as age-dependent and specific cut-offs for LA size by different age strata are not provided in neither the diastolic function recommendations nor the cardiac chamber quantification recommendations.1,2 As outlined below, this recommendation is based on several cross-sectional studies not showing age-related changes in LA volumes. Since our longitudinal findings, along with findings from the Multi-Ethnic Study of Atherosclerosis (MESA) study,14 suggest that LA volumes do increase with age, a reconsideration of this recommendation may be indicated. For other diastolic measures such as the e′, the diastolic function guidelines specifically emphasize that age should be taken into account when evaluating this measure.2 Furthermore, it has been proposed to tailor the diastolic function assessment by using age-appropriate cut-offs of the e′.15 The same could be considered for LA volumes since not providing age-stratified cut-offs may inadvertently designate elderly individuals as having abnormal LA size and consequently also diastolic dysfunction even though LA enlargement may simply reflect an age-related phenomenon.
Although data from the Framingham study did not show marked overall changes in the LA over time, that study evaluated LA dimension,16 which is a less sensitive marker than LA volume to detect remodelling of the LA.1 Based on cross-sectional studies, the LAVmax does not overall seem to increase with age,6,17 although some do suggest progression in late life, around the eighth decade.18 Our findings do, however, suggest that individuals without risk factors undergo remodelling of the LA, supporting that age-appropriate consideration of the LA may be warranted as noted above. One comparable report has investigated longitudinal changes in the LA as measured by cardiac magnetic resonance imaging (CMR) in MESA study (n = 2338).14 Although CMR is more accurate, the annual changes were comparable with ours. Furthermore, the MESA study only investigated participants with risk factors, and did therefore not determine progression in the absence of cardiac risk factors. Of note, they also found that LAEF decreased significantly during follow-up which is at odds with our findings. This may rely on the higher accuracy by CMR capable of detecting more minute changes. Furthermore, an overall change in LAEF is dependent on changes in active and passive LAEF, which change inversely with ageing.19 If these changes counteract each other, an absolute change in LAEF will not be observed, which may explain our observation.
Accelerators of LA remodelling
Although several clinical factors correlate with LA size, as we also observed in our study, the most influential determinants were age and AF at baseline, and only hypercholesterolemia was further independently significantly associated with LA remodelling. These findings are at odds with findings from the Framingham Study, which established several clinical factors to be associated with accelerated remodelling of the LA, including age, gender, BMI, blood pressure, and antihypertensive medication.16 Importantly, they excluded participants with known AF and did not investigate the impact of hypercholesterolemia, however, they do suggest more modifiable risk factors that could be targets for physicians in order to prevent or delay progression of LA remodelling. Although clinicians should seek to modify these risk factors, our study does not support that this will translate into less LA remodelling. Even though our finding that hypercholesterolemia accelerates LA remodelling could suggest that lipid-lowering treatment may potentially be of value, the reason for its selective impact on LAVmax is unclear. Accordingly, this finding needs further validation as it may be a chance finding. It is already well-established that rhythm management by both cardioversion and ablation treatment can result in reverse remodelling of the LA, and thereby potentially decrease the risk of clinical outcomes,20,21 hence AF rhythm management should also be a target for halting the progression of LA remodelling.
The reason for the discrepancy between our findings and the Framingham Study is unclear but may rely on the different LA parameters that were investigated as changes in LA dimension would likely signify substantial remodelling of the LA accelerated by clinical risk factors.
Of paraclinical factors, a notable finding was that proBNP was associated with change in LAVmin but not LAVmax. Furthermore, it showed a differential impact on LA remodelling according to presence of AF. Among those without AF, increasing natriuretic peptides signifies an abnormal LA endocrine function, whereas in those with AF the association between natriuretic peptides and LA function is more complex. In these subjects, a lower release of natriuretic peptides may signify an endocrine burnout due to existing LA remodelling and LA fibrosis.22 Hence, a higher level of natriuretic peptides may indicate a relatively healthy atrium with potential for reverse remodelling, which could explain these diverging findings according to presence of AF.
Prognostic value of LA remodelling
Even though it is well-established that larger LA size is associated with incident AF and HF,23 there are little data looking at the prognostic effects of longitudinal changes in LA measures. Based on a large-scale population cohort, the MESA study (n = 2338 with 132 AF events) also found that changes in LA volumes were predictive of incident AF, and added value beyond a clinical risk tool, the CHARGE-AF score.14 This is line with our findings. Similarly, since reverse remodelling after rhythm management is associated with a lower risk of AF recurrence,21 it is fair to assume that stagnation or progressive remodelling confers a higher risk of AF recurrence.
LA remodelling is a well-established predictor of HF,4 and progressive LA remodelling can be useful in tracking disease progression in HF.24 It is a frequently encountered observation in HF, typically as it enlarges secondary to structural abnormalities in the LV and secondary to separation of the subvalvular apparatus of the mitral valve leading to secondary mitral regurgitation.25,26 Although we are not aware of other studies showing that progressive LA remodelling is predictive of HF, it is likely that this association reflects the secondary effects from LV remodelling, and thereby LV disease progression. This is supported by the finding that LV size and LVMi were accelerators of LA remodelling. The selective impact of LVMi on changes in LAVmin, suggests that LAVmin remodelling is more sensitive to LV remodelling, likely as it is more closely linked to LV filling pressure.7 The selective impact of GLS on LAVmax may reflect that a primary determinant of LAVmax and LA reservoir function in general is the longitudinal displacement of the LV.
Progressive LA remodelling signals the cumulative burden of elevated filling pressure, which ultimately leads to pulmonary congestion, the hallmark of HF, explaining why it was a predictor of HF.25,27
Limitations
The baseline and follow-up investigations were performed with different ultrasound machines and different post-processing software. Although the ultrasound machines could provide differences in image quality, the post-processing software was not likely to influence the key measurements of LA volume quantification. Furthermore, two different investigators performed the LA volume measurements in the fourth and fifth examinations rounds, respectively. Even though these factors introduce potential inter-rater issues, these are often the necessary trade-offs when investigating long-term longitudinal changes with examinations performed more than a decade apart from each other.
Despite observing relatively few outcome events, we adjusted for several clinical and echocardiographic confounders. Although this could result in overfitting issues, this was done simply to stress the strengths of the associations between LA measures and outcomes.
Conclusion
In the general population, LA volumes increase significantly with age, which is not the case for the LAEF. These findings can also be appreciated in healthy individuals. Changes in LA volumes are mainly accelerated by age and AF, but also influenced by LV structure and function. LA remodelling poses an increased risk of subsequent AF and HF.
Data availability
Due to Danish legislation, we cannot make the data publicly available. Data output can, however, be obtained upon reasonable request.
Funding
F.J.O. was financed by the Danish Heart Foundation (Grant no. 18-R125-A8534-22083), Herlev & Gentofte Hospital’s Research Foundation, Kong Christian den Tiendes Fond, and Fru Asta Florida Boldings Mindelegat. The Copenhagen City Heart Study was funded by the Danish Heart Foundation, and the echocardiographic substudy was further funded by the Lundbeck Foundation.
Conflict of interest: T.B.-S.: Steering Committee member of the Amgen financed GALACTIC-HF trial; Steering Committee member of the Boston Scientific financed LUX-Dx TRENDS trial; Advisory Board: Sanofi Pasteur; Advisory Board: Amgen; Speaker Honorarium: Novartis; Speaker Honorarium: Sanofi Pasteur; and Research grant: GE Healthcare and Sanofi Pasteur.
The other authors have no potential conflicts of interest.
