OUP user menu

Effects of perindopril on cardiac remodelling and prognostic value of pre-discharge quantitative echocardiographic parameters in elderly patients after acute myocardial infarction: the PREAMI echo sub-study

Gian Luigi Nicolosi, Sorin Golcea, Claudio Ceconi, Giovanni Parrinello, Adriano Decarli, Massimo Chiariello, Willem J. Remme, Luigi Tavazzi, Roberto Ferrari
DOI: http://dx.doi.org/10.1093/eurheartj/ehp139 1656-1665 First published online: 30 April 2009


Aims To determine (i) the effect of perindopril on several geometric and functional parameters of the left and right ventricles assessed by echocardiography in the unique Perindopril and Remodelling in Elderly with Acute Myocardial Infarction (PREAMI) population of post-acute myocardial infarction (AMI) elderly patients with preserved left ventricular (LV) function; and (ii) the prognostic predictors at pre-discharge derived from echo-Doppler measurements in the same population.

Methods and results PREAMI included 1252 post-AMI patients (age 73 ± 6 years, LV ejection fraction 59.1 ± 7.7%) receiving optimal therapy after AMI, randomized to perindopril 8 mg/day (n = 631) or placebo (n = 621); n = 896 had complete echo-Doppler data. Outcome measures were clinical [death, heart failure (HF)] and standard echo-Doppler parameters. Pre-discharge LV end-diastolic volume (LVEDV) was similar: 81.1 ± 23.1 (perindopril) and 79.6 ± 22.7 mL (placebo). At 6 months and 1 year, LVEDV remained unchanged with perindopril (81.2 ± 24.4 and 81.8 ± 26.8 mL, respectively), but increased with placebo (83.0 ± 25.3 and 83.6 ± 25.7 mL, respectively, both P < 0.001 vs. baseline). Perindopril reduced cardiac sphericity vs. placebo (P = 0.015 at 6 months; P = 0.020 at 1 year). Classification regression tree analysis showed treatment as the most important predictor of remodelling. Multiple pre-discharge echocardiographic variables predicted the death/HF endpoint, independently of treatment (P ≤ 0.05).

Conclusion Remodelling occurs in post-AMI in elderly patients with normal LV function. Echo-Doppler variables at baseline have prognostic implications. Treatment with perindopril reduces progressive LV remodelling that can occur even in the case of small infarct size.

  • Echocardiography
  • Myocardial infarction
  • Perindopril
  • Prognosis
  • Remodelling


Acute myocardial infarction (AMI) can initiate progressive changes in left ventricular (LV) size, shape, mass, and geometry that impair LV function and, implicitly, prognosis.15 These changes, which are referred to as remodelling, may be acute or chronic and involve infarcted and non-infarcted areas.6 Remodelling is widely recognized as a precursor of cardiovascular events, especially heart failure (HF), and a strong predictor of mortality.1,3,7

In several large trials, angiotensin-converting enzyme (ACE)-inhibitors have been shown to improve prognosis and reduce remodelling and ischaemic events in selected patients with chronic LV dysfunction and HF811 or after AMI.12,13 In these studies, remodelling was usually measured in terms of changes in LV ejection fraction (LVEF) and ventricular diameters.

When we plot the mean LVEF of the patients involved in the various ACE-inhibitor trials against their age (Figure 1), it is clear that data on elderly patients (i.e. >65 years) with preserved LV function (i.e. LVEF > 40%) are missing. This is a relevant subset of patients, as the mean age of 50% of all post-AMI patients exceeds 65 years.14 In the Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto Miocardico-2 (GISSI-2) study, haemodynamic dysfunction and congestive HF were more severe in the elderly, despite similar infarct size across all age groups.15 The elderly are more prone to remodelling than younger patients16 and to HF with preserved LV systolic function,4 even after smaller infarcts.

Figure 1

Correlation between baseline mean left ventricular (LV) ejection fraction and mean age in trials of angiotensin-converting enzyme-inhibitors in coronary artery disease, including coronary artery disease prevention (black circles), heart failure (white circles), and acute myocardial infarction (grey circles). AIRE, Acute Infarction Ramipril Efficacy study; CAPTIN, captopril plus tissue plasminogen activator following acute myocardial infarction; CATS, Captopril and Thrombolysis Study; CONSENSUS, Cooperative New Scandinavian Enalapril Survival Study; EUROPA, European trial on Reduction of cardiac events with Perindopril in stable coronary Artery disease; FAMIS, Fosinopril in Acute Myocardial Infarction Study; GISSI-3, Gruppo Italiano per lo Studio della Sopravvivenza nell’Infarto miocardico; HOPE, Heart Outcomes Prevention Evaluation study; PEACE, Prevention of Events with Angiotensin-Converting Enzyme inhibition; PREAMI, Perindopril and Remodelling in Elderly with Acute Myocardial Infarction; QUIET, Quinapril Ischaemic Event Trial; SAVE, Survival and Ventricular Enlargement study; SOLVD, Studies of Left Ventricular Dysfunction.

Despite these observations, the dynamics of LV remodelling remain largely undocumented in the elderly, as is the long-term outcome of treatment with ACE-inhibition.5 The recent Perindopril and Remodelling in Elderly with Acute Myocardial Infarction (PREAMI) trial has contributed to fill this gap in knowledge. Treatment with perindopril reduced absolute risk of the combined primary endpoint of death, hospitalization for HF, and cardiac remodelling by 22% at 1 year (P < 0.001).17 Moreover, in PREAMI, LV remodelling, defined as an increase in LV end-diastolic volume (LVEDV) by ≥8%,17 was a component of the primary endpoint that was largely affected by active treatment with the ACE-inhibitor perindopril. Remodelling was present in 51.2% of the placebo group vs. 27.7% of the perindopril group (P < 0.001).

The first aim of the present study is to expand on the main results of PREAMI, by evaluating the effects of perindopril on the global (geometric and functional) remodelling of the LV and right ventricle (RV) assessed by central careful echo reading. The second aim is to determine the prognostic value of pre-discharge quantitative echo-Doppler measurements in this unique population of elderly post-AMI patients with preserved LV function (LVEF ≥ 40%) followed for 1 year.


Study design

The PREAMI study design has been described in detail elsewhere.4,17 Briefly, PREAMI was a double-blind, randomized, parallel group, multicentre study in 1252 patients 65 years or older who survived an AMI with preserved LVEF (≥40%) and optimal apical six-chamber views of the LV recorded for at least five complete cardiac cycles. The study was approved by the Ethics Committees of all participating centres, and conformed to the Declaration of Helsinki. Written informed consent was obtained from all patients.

Patients received either perindopril 8 mg/day (n = 631) or placebo (n = 621). The primary endpoint was a composite (in hierarchical order) of death, hospitalization for HF, and LV remodelling (considered as ≥8% increase in LVEDV measured by echo). Secondary endpoints included the primary objectives considered separately.

Echo-Doppler studies

Patients underwent three echo-Doppler studies, before discharge [median (inter-quartile range): 9 (4) days], at 6 months [197 (13) days], and at 1 year [382 (21) days] after AMI, each comprising multiple standard views on Super VHS 0.5 in. videotapes, as recommended by the American Society of Echocardiography (ASE).18 The echo-Dopplers were centrally assessed at the Core Echocardiographic Laboratory, Gussago (Brescia) for technical quality and eligibility for quantitative analysis. Minimal requirements were apical six-chamber views that allowed proper calculation of LV volumes and maximized LV cavity size (avoiding foreshortening) with good endocardial border visualization. Wall motion analysis was performed using a 16-segment model of the LV. The three investigators blindly assigned consensus segment scores using the semi-quantitative ASE visual grading system.18 Wall motion score index (WMSI) was derived from the sum of segmental scores divided by the number of segments. Dyssynergy [percentage of wall motion abnormalities (% WMA)], a crude marker of ischaemic damage, was quantified by dividing the number of akinetic, dyskinetic, and aneurysmal segments (segmental score ≥3) by total segments assessed × 100.5 Documented intraobserver and interobserver reproducibilities were 93 and 89%.

Selected images were digitized (TomTec Medical Imaging off-line analysis system) to provide ventricular endocardial contours and measure LVEDV and LV end-systolic volume (LVESV) from two apical orthogonal views (six-chamber views). Papillary muscles were included within the cavity. The modified Simpson’s rule was used to calculate biplane LV volumes and LVEF.18 All measurements were obtained blind to treatment from three cardiac cycles and averaged for further analysis.

The interobserver variability for LVEDV and LVESV, measured by the intraclass correlation coefficient in 76 random patients, was 2.0% [95% confidence interval (CI): 1.3–3.1] and 1.5% (95% CI: 0.9–2.3), respectively, and the intraobserver variability was 1.0% (95% CI: 0.7–1.7) and 0.7% (95% CI: 0.5–1.1), consistent with previous studies.5,19

The LV shape was assessed using the non-dimensional, non-linear eccentricity index (EI) of sphericity, from 0 (perfect sphere) to 1 (increasing linearity).18,20 The radius–wall thickness ratio was used as a guide to detect LV hypertrophy (i.e. the ratio of the LV cavity radius at mid-ventricle to the interventricular septum thickness at end-diastole measured from the parasternal long-axis view 1.5 cm below the aortic ring).18

Mitral regurgitation (MR) was semi-quantitatively assessed as mild, moderate, or severe by colour Doppler flow mapping,21 with 98% interobserver concordance. GISSI-3 procedures were followed for end-systolic left atrial (LA) planimetry22 and measurement of RV systolic function on the basis of tricuspid annular plane systolic excursion (TAPSE).23 LV diastolic filling patterns were obtained from pulsed Doppler mitral inflow recordings in the apical four-chamber view sample volume 1 cm below the mitral annular plane between the mitral leaflet tips. LV diastolic function was obtained from mitral Doppler tracings by peak flow velocity of early filling (E wave), peak flow velocity of atrial contraction (A wave), the E/A ratio, and early filling deceleration time (DT), averaged from three consecutive cycles in sinus rhythm or five consecutive cycles in atrial fibrillation.4,19 The Valsava manoeuvre was used to unmask pseudonormalization. The E/A ratio and DT were used to assign patients to a either (i) a pseudonormal/restrictive group, defined by E/A ≥ 1.6 or DT ≤ 150 ms or (ii) a non-restrictive group, defined by E/A < 1.6 or DT > 150 ms. These cutoff values were selected on the basis of age-adjusted means ± 2 SD, as described elsewhere.24 Clinical endpoints were adjudicated by an independent review committee, as previously described.4,17


Descriptive analysis

Descriptive statistics are expressed as means ± SD and/or as median and inter-quartile ranges, according to distribution. Groups were compared using Student’s t-test or the non-parametric Mann–Whitney test for continuous variables, with the χ2 for the categorical tests. Correlations between continuous variables were evaluated using Pearson’s or Spearman’s correlation coefficient. Non-normally distributed parameters were suitably transformed [EDV, ESV, and LA area were log-transformed, while for EF and EI in systole (four chambers) the arcsin transformation was applied].

Linear mixed model

Time and treatment variations in echo-Doppler parameters were tested using a linear mixed model to take into account the intrasubject correlation of measures. Prognostic factors for remodelling at 1 year were estimated by univariable and multivariable logistic regression. In the mixed model, we used each patient as a random factor and the time, treatment, and baseline parameters as fixed factors. We applied Diggle’s approach to determine the best parsimonious model.25 Moreover, we selected best covariance structure by means of the method recommended by Kincaid.26

Classification and regression tree and multivariable analysis

Prognostic factors for remodelling at 1 year were estimated by univariable and multivariable logistic regression using demographics (sex, age), clinical history (previous AMI, hypertension, stroke/transient ischaemic attack, diabetes), co-treatments (statins, β-blockers, nitrates, diuretics), echo parameters, and treatment. The significant variables from the univariable model were used in the multivariable analysis. In selecting variables for multivariable analysis, allowance was made for intervariable correlation to avoid colinearity. The same parameters were tested as event predictors (death or HF) using univariable analysis (log-rank test) and multivariable analysis (Cox proportional hazard). Verification of proportional hazard assumptions was performed by Schoenfeld residuals.27 Furthermore, in selecting significant variables for multivariable analysis, allowance was made for intervariable correlation to avoid colinearity and unstable models (e.g. EDV and ESV, which had a correlation of r = 0.9). The same parameters were tested as predictors of events (death, HF, or both) using univariable analysis (log-rank test) and multivariable analysis (Cox proportional hazards). A Martingale residual plot was used to evaluate the proportionality assumption.

The Akaike information criterion was used to select the best model. The final model was validated using two approaches: (i) the C-index to measure the discrimination ability and (ii) the Hosmer–Lemeshow test for its calibration. Classification and regression tree analysis was performed to test the strength of the analytical conclusions, i.e. to reveal possible predictive interactions and develop a risk stratification model. The pattern of the relationship between continuous predictors and remodelling risk and/or survival was studied using restricted cubic spline regressions. To control for the type I error related to multiple comparisons, we applied the approach of Hochberg and Tamhane.28

Validation of the final model

Model performance was evaluated by measures of calibration (Hosmer–Lemeshow statistics) and discrimination. Discrimination refers to the ability of a model to assign higher probabilities of death (outcome) to patients who actually die than those patients who live. This was evaluated by the area under the receiver operating characteristic (ROC) curve, which is equivalent to the C-index. The C-index derived from the multivariable models was used to assess the improvement in the prognostic model discrimination resulting from the sequential addition of each echo parameters to a model including clinical risk factors, selected in the PREAMI study. This statistic was calculated after having applied a resampling validation by bootstrap. Comparisons between the areas under the ROC curves were performed by pairwise method with the use of U statistics. ROC curves for time-dependent outcomes were also calculated.

The R language statistical package was used, and tests were two-sided with P < 0.05 throughout.29

To test possible selection bias related to deaths or missing echo data, we performed two additional analyses: (i) a generalized (logistic) mixed model with missing data (Y/N) at the three visits as dependent variables to evaluate different trends in the intermittent missingness processes between treatments and (ii) a logistic regression model to evaluate the predictors of remodelling by not excluding the deaths, but considering them as remodelling. Neither model showed a relationship between treatment group and missingness.


The mean age of the total population (n = 1252) was 73 ± 6 years; other baseline characteristics are presented in Table 1.17 The pre-discharge echo-Doppler parameters of the two groups (Table 2) show that the E/A ratio was the only variable that significantly differed (P = 0.01), though this is clinically irrelevant and possibly due to chance. All patients received optimal therapy following AMI, including thrombolysis and revascularization, according to local facilities: 533 (43%) patients received thrombolysis and 51 (4%) received primary angioplasty; 15 patients were shifted during follow-up to mechanical revascularization after thrombolysis. Table 3 shows concomitant therapy, which increased markedly after AMI and remained as such until the end of the study.17

View this table:
Table 1

Baseline clinical and echo-Doppler parameters per treatment group

CharacteristicsPerindopril (n = 631)Placebo (n = 621)P-value
 Age (years, mean ± SD)72 ± 673 ± 50.07
 Female, n (%)219 (35)217 (35)0.95
Clinical variables
 Weight (kg, mean ± SD)73 ± 1172 ± 110.57
 Body mass index (kg/m2, mean ± SD)27 ± 426 ± 40.43
 Heart rate (b.p.m., mean ± SD)67 ± 1067 ± 110.87
 SBP (mm Hg, mean ± SD)126 ± 15125 ± 140.80
 DBP (mm Hg, mean ± SD)74 ± 974 ± 90.19
 Current heart failure class, n (%)0.18
  NYHA I483 (77)501 (81)
  NYHA II140 (22)112 (18)
  NYHA III8 (1)7 (1)
Previous medical history, n (%)
 Myocardial infarction66 (11)62 (10)0.22
 Revascularization (PTCA or CABG)19 (3)20 (3)0.82
 Heart failure11 (2)7 (1)0.18
 Hypertensiona373 (59)355 (57)0.44
 Stroke or TIA35 (6)42 (7)0.15
 Peripheral vascular disease49 (8)37 (6)0.40
 Diabetes mellitus157 (25)143 (23)0.48
 Hyperlipidaemiab175 (28)182 (29)0.63
 Smokers266 (42)256 (41)0.98
 Current smokers122 (19)120 (19)0.85
Previous medications, n (%)
 Antithrombotics117 (19)114 (18)0.93
 β-blockers82 (13)69 (11)0.31
 Lipid-lowering agents62 (10)60 (10)0.92
 Nitrates80 (13)81 (13)0.85
 Calcium channel blockers134 (21)154 (25)0.13
 Diuretics106 (17)91 (15)0.08
 ACE-inhibitorsc154 (24)149 (24)0.86
 Thrombolysis given272 (43)261 (42)0.99
Biological parameters
 CPK release (U/L, mean ± SD)1249 ± 10581237 ± 10460.85
  • ACE, angiotensin-converting enzyme; CPK, creatine phosphokinase; CABG, coronary artery bypass graft; DBP, diastolic blood pressure; NYHA, New York Heart Association; PTCA, percutaneous transluminal coronary angioplasty; SBP, systolic blood pressure; SD, standard deviation; TIA, transient ischaemic attack.

  • aKnown history of hypertension or receiving antihypertensive treatment.

  • bKnown history of hyperlipidaemia or receiving lipid-lowering therapy.

  • cACE-inhibitors were withdrawn at least 24 h before randomization.

View this table:
Table 2

Echo-Doppler profiles per treatment group

Echo-Doppler parameterPre-discharge6 months1 year
Perindopril (n = 550)Placebo (n = 538)Perindopril (n = 465)Placebo (n = 453)Perindopril (n = 476)Placebo (n = 460)
LVEDV (mL)*81.1 ± 23.179.6 ± 22.781.2 ± 24.4***83.0 ± 25.3**81.8 ± 26.8***83.6 ± 25.7**
LVESV (mL)*34.2 ± 14.333.1 ± 14.334.0 ± 16.5***34.9 ± 16.7**34.6 ± 17.935.1 ± 17.3**
LVEF (%)*58.9 ± 7.759.3 ± 7.859.4 ± 8.459.2 ± 8.159.0 ± 8.659.4 ± 8.7
WMSI1.29 ± 0.261.29 ± 0.271.22 ± 0.261.23 ± 0.27**1.19 ± 0.221.23 ± 0.28**
Wall motion asynergy (%)*12.2 ± 12.612.7 ± 12.49.0 ± 12.2**9.1 ± 11.6**8.0 ± 11.2**8.5 ± 10.7**
EI in diastole (four chambers)0.73 ± 0.090.74 ± 0.090.73 ± 0.090.72 ± 0.08**0.73 ± 0.090.73 ± 0.08
EI in systole (four chambers)0.82 ± 0.090.82 ± 0.080.82 ± 0.07***0.81 ± 0.08**0.81 ± 0.08***0.80 ± 0.09**
Radius–wall thickness ratio1.92 ± 0.381.88 ± 0.361.92 ± 0.381.88 ± 0.361.92 ± 0.381.88 ± 0.36
LA area (cm2)*21.7 ± 4.218.1 ± 4.020.9 ± 5.018.4 ± 3.920.5 ± 5.018.5 ± 4.1**
Mitral regurgitation, n (%)329 (59.8)332 (61.7)263 (56.6)254 (56.1)283 (59.5)273 (59.3)
TAPSE (cm)1.69 ± 0.451.81 ± 0.441.79 ± 0.501.79 ± 0.451.77 ± 0.451.78 ± 0.44
E/A ratio0.92 ± 0.40***0.97 ± 0.380.91 ± 0.490.94 ± 0.71**0.87 ± 0.36**0.91 ± 0.46**
DT (ms)239 ± 58240 ± 58256 ± 58**260 ± 66**258 ± 61**260 ± 62**
  • Values are presented as mean ± SD, unless otherwise specified [n (%)]. DT, deceleration time; E/A, ratio of E wave to A wave; EI, eccentricity index; LA, left atrium; LVEDV, left ventricular end-diastolic volume; LVEF, left ventricular ejection fraction; LVESV, left ventricular end-systolic volume; SD, standard deviation; TAPSE, tricuspid annular plane systolic excursion; WMSI, wall motion score index.

  • *P < 0.05 remodelling-induced behaviour.

  • **P < 0.05 within-group vs. baseline.

  • ***P < 0.05 between groups (perindopril-added effect).

View this table:
Table 3

Concomitant therapy17

TherapyPerindopril n (%)Placebo n (%)
Before AMI (n = 631)After AMI (n = 631)At 12 months (n = 528)Before AMI (n = 621)After AMI (n = 621)At 12 months (n = 521)
Anti-thrombotics120 (19%)618 (98%)469 (94%)112 (18%)609 (98%)485 (93%)
β-Blockers82 (13%)442 (70%)343 (65%)68 (11%)447 (72%)344 (66%)
Lipid-lowering agents63 (10%)309 (49%)285 (54%)62 (10%)323 (52%)292 (56%)
Nitrates82 (13%)517 (82%)338 (64%)81 (13%)509 (82%)339 (65%)
Calcium channel blockers133 (21%)120 (19%)143 (27%)155 (25%)137 (22%)167 (32%)
Diuretics107 (17%)177 (28%)148 (28%)93 (15%)161 (26%)135 (26%)
ACE-inhibitorsa151 (24%)486 (77%)42 (8%)149 (24%)491 (79%)42 (8%)
  • ACE, angiotensin-converting enzyme; AMI, acute myocardial infarction.

  • aACE-inhibitors were withdrawn at least 24 h before randomization.

  • Reproduced from The PREAMI Investigators17 with permission.

Complete sets of echocardiographic data were unavailable for 356 patients, due to loss to follow-up, refusal, tape damage, courier loss, or poor echocardiographic quality. Only one patient was lost to follow-up. The final number of data missing was as anticipated and similar to previous studies.5,6 Validated sets of predischarge and follow-up (6 months and 1 year) echo-Doppler data were analysed for 455 patients in the perindopril group and 441 patients in the placebo group.17 This population did not differ from the entire sample for any clinical or laboratory parameter.

Effect of perindopril on echo-Doppler variables

Predischarge LVEDV values were similar in the two groups (Table 2). At 6 months and 1 year, LVEDV remained unchanged with active treatment, but significantly increased with placebo (P < 0.001 vs. baseline). Mean increases in LVEDV were 0.7 mL with perindopril vs. 4.0 mL with placebo (P < 0.001 vs. baseline). The difference between the active treatment and placebo groups was significant at 6 months (1.8 mL, P = 0.007) and 1 year (1.8 mL, P = 0.003). The changes in LVESV followed a similar pattern, but were smaller, and the difference between perindopril and placebo was significant at 6 months only (P = 0.042, Table 2).

Perindopril did not affect LVEF, WMSI, WMA, EI in diastole, radius–wall thickness ratio, LA area, TAPSE, E/A ratio, DT, or restrictive pattern. The LVEF of the overall population was normal at baseline (59.1%) and remained as such for the entire duration of the study with no between-group differences.

Pre-discharge WMA and WMSI were small, and significantly decreased during follow-up (P < 0.001 for WMA in the whole population), independently of treatment (Table 2). In contrast to EI in diastole, EI in systole progressed toward sphericity in both groups (P < 0.001), a phenomenon significantly attenuated by perindopril at 6 months and 1 year (P = 0.015 and P = 0.020 vs. placebo, respectively; Table 2).

Mean DT increased significantly in the pseudonormal/restrictive group and became comparable to that of the non-restrictive group over 1 year (Table 4). The number of pseudonormal/restrictive patients declined by 20% during follow-up (Figure 2). The ischaemic area in the pseudonormal/restrictive group was larger (WMA, 15.4 vs. 12.0%, P < 0.0147), LVEF was lower, and LV systolic shape was more spherical than in the non-restrictive group [EI in systole declined from 0.82 to 0.79 (P = 0.001) in the pseudonormal/restrictive group vs. 0.82 to 0.81 (P = 0.001) in the non-restrictive group].

Figure 2

Changes in left ventricular filling patterns after acute myocardial infarction over 1-year follow-up in 678 out of 1017 patients with complete echocardiographic data, and distribution of patients between non-restrictive and pseudonormal/restrictive groups. DT, deceleration time; E/A, ratio of E wave to A wave.

View this table:
Table 4

Deceleration time (ms) at pre-discharge and after 6 months and 1 year in patients in the pseudonormal/restrictive group (E/A ≥ 1.6 or DT ≤ 150 ms) vs. those in the non-restrictive group (E/A< 1.6 or DT > 150 ms) with pre-discharge Doppler data (n = 1017)

Deceleration time (ms, mean ± SD)Within-group significance
Pre-discharge6 months1 year
Non-restrictive group (n = 930)246 ± 56263 ± 60262 ± 59P = 0.001
Pseudonormal/restrictive group (n = 87)188 ± 54224 ± 65228 ± 58P = 0.01
Between-group significanceP = 0.46P = 0.016P = 0.001
  • E/A, ratio of E wave to A wave.

Predischarge echo-Doppler predictors of clinical events

Death or HF at 1 year occurred in 120 out of 1252 patients (9.6%; n = 55 perindopril; n = 65 placebo). There were 39 deaths (6.2%) in the perindopril group and 16 hospitalizations for HF (2.5%) vs. 38 (6.1%) and 27 (4.3%) in the placebo group.17

Univariable survival analysis showed that pre-discharge LVEDV, LVESV, LVEF, EI in diastole, and DT all predicted 1-year mortality (P ≤ 0.05). In the multivariable analysis, LVEDV, LVESV, LVEF, and MR maintain their predictive role for 1-year mortality. Only pre-discharge LVEDV, LVESV, and LA area independently predicted HF (P ≤ 0.05; Table 5). Pre-discharge LVEDV, LVESV, LVEF, WMA, and LA area predicted death and HF (P ≤ 0.05; Table 5). As shown in Figure 3, there was a linear relationship between death and HF and LVEF for values between 40 and 60%. For values between 60 and 70%, the relationship was no longer linear and the curve flattened out. As regards the development of remodelling, this was predicted only by pre-discharge LVEDV, LVESV, and WMA; these parameters maintained their predictive capacity in the multivariable approach.

Figure 3

Risk of composite event [death and hospitalization for heart failure (HF)] according to baseline left ventricular (LV) ejection fraction. Dotted lines show 95% confidence intervals. Test for linearity, P = 0.06.

View this table:
Table 5

Prognostic significance of pre-discharge echo-Doppler parameters

Echo-Doppler parameterDeath (n = 77)HF (n = 52)Death + HF (n = 120)Remodelling (n = 352)
Univariable analysis
 LVEDVP = 0.004P = 0.001P < 0.001P < 0.001
 LVESVP < 0.001P = 0.02P < 0.001P < 0.001
 LVEFP < 0.001P = 0.67P < 0.01P = 0.30
 WMSIP = 0.25P = 0.36P < 0.01P = 0.75
 Wall motion asynergyP = 0.07P = 0.12P = 0.18P = 0.02
 EI in diastole (four chambers)P = 0.04P = 0.22P = 0.71P = 0.09
 EI in systole (four chambers)P = 0.10P = 0.59P = 0.71P = 0.38
 Radius–wall thickness ratioP = 0.32P = 0.80P = 0.18P = 0.14
 LA areaP = 0.14P < 0.01P < 0.01P = 0.21
 Mitral regurgitationP = 0.04P = 0.55P = 0.09P = 0.83
 TAPSEP = 0.78P = 0.49P = 0.25P = 0.99
E/AP = 0.39P = 0.90P = 0.70P = 0.56
 DTP = 0.04P = 0.62P = 0.87P = 0.42
Multivariable analysis, hazard ratio (95% confidence interval)
 LVEDV1.02 (1.01–1.03), P = 0.0041.02 (1.01–1.04), P = 0.0011.02 (1.01–103), P = 0.0010.97 (0.96–0.99), P = 0.001
 LVESV1.03 (1.01–1.05), P = 0.0011.03 (1.00–1.05), P = 0.0231.03 (1.01–1.04), P = 0.0010.97 (0.96–0.98), P = 0.001
 LVEF0.94 (0.91–0.97), P = 0.0010.99 (0.95–1.04), P = 0.670.96 (0.93–0.99), P = 0.0041.01 (0.99–1.03), P = 0.30
 Wall motion asynergy1.02 (0.99–1.04), P = 0.0661.03 (1.01–1.06), P = 0.011.02 (1.01–1.04), P = 0.0041.01 (1.00–1.03), P = 0.024
 EI in diastole (four chambers)0.28 (0.02–4.87), P = 0.3914.13 (0.20–986), P = 0.221.61 (0.13–20.42), P = 0.714.81 (0.77–25.91), P = 0.094
 Radius–wall thickness ratio0.67 (0.30–1.48), P = 0.320.88 (0.33–2.34), P = 0.800.57 (0.32–0.98), P = 0.0450.71 (0.46–1.11), P = 0.14
 LA area1.05 (0.98–1.11), P = 0.141.12 (1.04–1.19), P = 0.0021.07 (1.02–1.27), P = 0.0030.98 (0.94–1.01), P = 0.21
 Mitral regurgitation0.76 (0.59–0.99), P = 0.0390.91 (0.65–1.26), P = 0.550.84 (0.68–1.03), P = 0.0891.02 (0.86–1.21), P = 0.80
 TAPSE0.90 (0.43–1.89), P = 0.780.74 (0.31–1.75), P = 0.490.71 (0.39–1.28), P = 0.2501.00 (0.64–1.56), P = 0.997
 DT0.27 (0.002–38), P = 0.613.82 (0.02–726), P = 0.620.74 (0.02–32.52), P = 0.872.93 (0.21–41.08), P = 0.42
  • DT, deceleration time; E/A, ratio of E wave to A wave; EI, eccentricity index; LA, left atrium; LVEDV, left ventricular end-diastolic volume; LVEF, left ventricular ejection fraction; LVESV, left ventricular end-systolic volume; TAPSE, tricuspid annular plane systolic excursion; WMSI, wall motion score index.

  • Multivariable analysis is adjusted for treatment, hypertension, sex, age, diabetes, diastolic blood pressure, systolic blood pressure, body mass index, heart rate, peripheral vascular disease, previous acute myocardial infarction, cerebrovascular disease, hyperlipidaemia, smoking, previous heart failure. Only Echo-Doppler parameters with a statistical significant effect in the univariable analysis on at least one endpoint were included.

Classification and regression tree analysis, using the clinical and echo-Doppler parameters, as defined in the Statistics section, confirmed the positive influence of treatment on remodelling, with a discriminating effect of LVEDV (cutoffs differed between treatment arms). Dyssynergy predicted remodelling only with LVEDV > 57.52 mL (Figure 4). This analysis may be useful in predicting cutoffs in clinical practice.

Figure 4

Classification and regression tree analysis for prediction of risk of remodelling according to treatment (perindopril or placebo), left ventricular end-diastolic volume (LVEDV), and dyssynergy. Risk represents the risk of remodelling. *Level of significance for the choice of threshold.


Elderly post-AMI patients with normal LVEF undergo progressive LV enlargement between discharge and 1 year mimicking that is seen in younger patients with larger infarcts and greater LV dysfunction.3,6,30,31 This may be surprising since these elderly patients could be considered to be at relatively low risk because of their normal LVEF. This suggests that age, in addition to infarct size, is one of the determinants of remodelling.17,19

Our data also show that the elderly with smaller volumes and greater dyssynergy are more prone to progressive remodelling. Small LV cavity size is common in the elderly due to concomitant hypertension and concentric LV hypertrophy; 58% of the PREAMI population had a history of hypertension.17 In addition, the radius–wall thickness ratio was lower in the remodelling subgroup compared with the overall population (1.84 vs. 1.90, P = 0.033). As expected, there was a treatment-independent recovery of WMA during the follow-up. This is most likely due to the occurrence of stunning and/or hibernation, as described in younger patients.5,6,32,33 We did not study elderly patients at high risk of remodelling complications because this population is already well studied.

Perindopril prevented remodelling by reducing LV dilatation. The majority of the effect was already evident after 6 months and was similar to that obtained with ACE-inhibition in younger patients with LV dysfunction.30,31 Perindopril did not affect LVEF or LV diastolic shape, though treatment improved the geometry of LV contraction, as expressed by the EI in systole.20,34 All these effects of perindopril occurred, despite concomitant use of β-blocker in 71% of patients, with no interaction among treatments.17

Several echo-Doppler parameters, not all LV-related, predicted death and/or HF at 1 year. This reinforces the concept of global cardiac remodelling and confirms the prognostic utility of complete pre-discharge echocardiography in elderly post-AMI patients. Our analysis identified four independent predictors of separate or combined 1-year endpoints: LVEDV, LVESV, WMA, and LA area (Table 5). Left ventricular EF was a linear predictor of the combined risk at values <60%, as reported in younger patients with moderate-to-severe LV dysfunction, for whom the relationship is more hyperbolic (Figure 3).2 The inflection points on the smoothed plots of radius–wall thickness ratio and TAPSE against combined event risk emphasize the importance of normal LV geometry and preserved RV systolic function.

Echo-Doppler-derived diastolic filling variables are important predictors of cardiac mortality in symptomatic and asymptomatic LV systolic dysfunction.19,3537 The restrictive pattern was reported as the best predictor of post-AMI cardiac mortality,35 adding significant prognostic information to established indicators of systolic dysfunction and identifying a subset at maximal risk. Our data suggest that restrictive filling also helps stratify combined event risk in elderly with normal-to-moderate systolic dysfunction.

Mean LVESV was larger in our pseudonormal/restrictive group and remained so during follow-up, suggesting impaired myocardial contraction rather than wall hypertrophy. Other severity criteria in this pseudonormal/restrictive subgroup were greater ischaemic area, lower LVEF, and a trend towards LV systolic sphericity during the follow-up.

One limitation to our study was that, though 2D echocardiography is a standard investigation of cardiac morphology and function,5,13 3D echocardiography and magnetic resonance imaging may be more accurate. However, assuming random measurement error, the use of 2D echocardiography cannot account for statistically significant echocardiographic changes. Another limitation is that, because the study was performed in multiple centres with differing device capabilities, it was not feasible to further quantify diastolic function using tissue Doppler imaging. E/A ratio and DT as diastolic pattern discriminators may appear simplistic, but they are well-established parameters.24,35 We should also note that 25–30% of the patients were lost to echocardiographic follow-up; however, this loss is usual in an elderly population and was foreseen in the statistical model. It is important to notice that remodelling is a dynamic process occurring over time. Therefore, in addition to the analysis between the two groups, we have also undertaken a longitudinal analysis within groups, which adds further information in this particular clinical setting, which was never studied before. Finally, the PREAMI sample size and duration produced lower mortality and HF hospitalization rates than expected (4% recorded vs. 14% anticipated).17 Therefore, larger and, particularly, longer trials are required to generate data on perindopril’s benefit on mortality and HF in such a population.

Clinical implications and conclusions

The data of this sub-study expand the clinical implication of PREAMI by demonstrating that LV remodelling can occur in elderly patients with AMI, despite small infarct size and optimized treatment.

Complete echo-Doppler workup should be mandatory for risk stratification in elderly post-AMI patients. Low volumes, marked asynergy, and pseudonormal/restrictive diastolic patterns need individualized, more aggressive management, particularly regarding remodelling. Addition of 8 mg/day perindopril tert-butylamine to optimized therapy significantly counteracts progressive LV dilation and improves cardiac sphericity. This therapy is well tolerated and should be considered as standard treatment in this ever-increasing clinical setting, for whom therapeutic decision-making is still equivocal and incompletely defined in current practice guidelines.


This study was supported by Stroder, Florence, Italy, and Servier Italia, Rome, Italy.

Conflict of interest: R.F., M.C., G.L.N., W.J.R., and L.T. have received honoraria from Salvatore Maugeri Foundation on behalf of the Sponsor.


View Abstract