Takotsubo syndrome is an increasingly recognized cause of chest pain and occasionally of cardiogenic shock. Despite rapid improvement of the left ventricular (LV) ejection fraction, recent registry data raise concerns about long-term prognosis. The aim of this study was to test the hypothesis that restoration of normal ejection fraction after acute takotsubo syndrome is not equivalent to full functional recovery.
Fifty-two patients with takotsubo syndrome (according to the Mayo Clinic criteria plus cardiac magnetic resonance imaging to exclude myocardial infarction) and 44 healthy control subjects of the same age, gender, and cardiovascular comorbidity distribution were prospectively recruited. The focus of the investigation was on patients with takotsubo syndrome presenting with ST-segment elevation–type electrocardiographic findings or malignant arrhythmias and with LV apical ballooning variant, and a 4-month recovery endpoint was assessed. Patients underwent echocardiographic assessment of LV myocardial deformation (global longitudinal, radial, and circumferential strain; LV twist, torsion, and untwist; and time to peak twist and untwist) and assessment of LV myocardial structure by pre- and post-contrast-enhanced cardiac magnetic resonance by T1 mapping acutely and at 4-month follow-up. Control subjects underwent a single-time-point investigation. Data were analyzed using paired or unpaired tests, as appropriate for their distribution, and corrected for multiple comparisons.
The patients’ mean age was 66 years (range, 28–87 years), and 92% were women. All abnormal echocardiographic indices observed acutely in patients with takotsubo syndrome improved (but did not necessarily normalize) at follow-up. Significant mechanotemporal alterations characterizing both systole (global longitudinal strain and apical circumferential strain, P < .01 for both; LV twist, twist rate, and torsion, P < .0001 for all) and diastole (untwist rate and time to peak untwisting, P < .001 for both) persisted at 4-month follow-up compared with control subjects, despite normalization of LV ejection fraction and volumes. Although native T1 (which demonstrates edema) normalized at 4-months follow-up only in segments contracting normally during the acute phase (T1 = 1,180 ± 40.6 msec [normally contracting segments, P = .20 vs control value of 1,189 ± 16 msec] and T1 = 1,208 ± 60.3 msec [dysfunctional segments, P < .05 vs control]), the extracellular volume fraction (which demonstrates diffuse fibrosis) remained significantly abnormal in all LV segments (whether normally contracting [0.328 ± 0.043, P < .001] or ballooning during acute presentation [0.320 ± 0.044, P < .001], both vs control value of 0.273 ± 0.045).
In patients with the most clinically severe spectrum of takotsubo cardiomyopathy, regional LV systolic and diastolic deformation abnormalities persist beyond the acute event, despite normalization of global LV ejection fraction and size. In addition, although myocardial edema partly subsides, a process of global microscopic fibrosis develops in its place, detected as early as 4 months.
Acute stress-induced (takotsubo) cardiomyopathy has become much more readily recognized in recent years. Patients with takotsubo cardiomyopathy, which commonly masquerades as acute myocardial infarction, often have emotional or physical triggers and show no evidence of a coronary culprit or infarct-like myocardial scar on cardiac magnetic resonance (CMR) imaging. The largest takotsubo registry in the world as well as data from the Swedish Cardiac Catheterisation Laboratory Registry demonstrate unequivocally that the medium- and long-term prognosis is not benign, and the mortality rate of takotsubo syndrome is comparable with that of myocardial infarction. Although this introduces a paradigm shift in how takotsubo syndrome’s prognosis was viewed until recently, new important questions have arisen, with an urgent need to explain why a condition considered until recently as “transient” and “naturally recovering” portends such a poor outlook. It has become increasingly obvious that the relatively rapid recovery of left ventricular ejection fraction (LVEF), with or without right ventricular ejection fraction, after an acute attack of takotsubo cardiomyopathy is not paralleled by symptomatic or electrocardiographic recovery, for example. We have previously shown that acute takotsubo hearts demonstrate severe global edema associated with profound cardiac energetic impairment, with incomplete resolution of both at 4 months ; additionally, we showed persistently abnormal global longitudinal strain (GLS) at 3-month follow-up in a cohort of 36 patients with takotsubo cardiomyopathy comprising all comers (ST-segment elevation, non–ST-segment elevation, normal electrocardiographic results, and all left ventricular [LV] ballooning subtypes [i.e., apical, midcavity, and basal]). However, in a condition that shows that it is on its way to recovery on the basis of several clear objective assessments (normalized ejection fraction, decrease of myocardial edema), there remains a disproportionately high rate of persistent symptoms, as well as lack of return to work and physical activity, which do not appear to improve at the same rate as the objective indices. Because edema, energetic impairment, and an isolated reduction in GLS do not preclude the coexistence or recovery of a normal resting LVEF (which is different from LV performance), there must be additional pathophysiologic processes that intervene and, if present in the long term, could begin to explain the unexpected poor outcomes of these patients. Discrepancies in previous literature may have arisen because of a wide clinical spectrum of patients with takotsubo syndrome, variable follow-up, or both. The only way to further test the hypothesis of remaining uncovered pathology to explain this disproportionate persistence of symptoms relative to apparent recovery is to focus the investigation on a homogeneous and most clinically severe presenting spectrum of patients with takotsubo cardiomyopathy (ST-segment elevation, malignant arrhythmias on presentation, and large, apical extending into midcavity ballooning phenotype) and to look into all (systolic and diastolic) parameters of functional nonrecovery and mechanisms of potential longer lasting changes, such as myocardial fibrosis. Herein, we describe the 4-months follow-up recovery of LV myocardial deformation and structure in a prospectively recruited cohort of patients with takotsubo syndrome presenting with ST-segment elevation, left bundle branch block or malignant arrhythmias on electrocardiography, and LV apical ballooning subtype.
Consecutive consenting patients who met the takotsubo Mayo Clinic and currently updated criteria (pheochromocytoma and myocarditis excluded also) and who, in addition, demonstrated no evidence of late gadolinium enhancement on CMR imaging were prospectively recruited at Aberdeen Royal Infirmary between 2011 and 2016. All patients were initially suspected of having myocardial infarction, and most were brought in by Scottish Ambulance Service for primary percutaneous coronary intervention. All patients had either an emotional or no clear trigger, and we purposefully excluded patients with other concomitant physical triggers that could have confounded our findings; thus, patients with infections, inflammatory conditions, or any other concurrent physical illnesses potentially influencing myocardial findings in our judgement were not included in this report. All patients underwent coronary angiography and left ventriculography and had serial electrocardiography, 12-hour blood troponin I and brain natriuretic peptide assessment, and comprehensive echocardiography and contrast-enhanced CMR within 5 days of the onset of symptoms (the mean time from symptoms onset to first echocardiographic examination was 2 days and to first CMR was 3 days). LV recovery was assessed using comprehensive echocardiography and CMR at 4-month follow-up (mean, 122 ± 5 days). To account for the effect of potential confounders, a control group of 44 subjects of similar age, gender, and comorbidity distribution (matching hypertension) were recruited and underwent a single-time-point echocardiographic assessment. The study was approved by the North of Scotland Research Ethics Committee and ran in accordance with the Declaration of Helsinki, and all subjects provided written informed consent.
A comprehensive two-dimensional echocardiographic protocol was performed using commercially available systems (Vivid E9, equipped with a 2.5-MHz [M5S] transducer; GE Vingmed Ultrasound, Horten, Norway) by the same operator in all subjects (J.S.). All measurements were taken from the first echocardiographic study performed at presentation and the designated follow-up study at 4 months. Cine loops in standard parasternal long-axis, short-axis, and apical four-, three-, and two-chamber views were obtained at a frame rate of ≥85 Hz. Short-axis views of the left ventricle were taken at the mitral valve level (tip of the mitral valve leaflets), midventricular level (papillary muscle), and apical level (absence of papillary muscles). Three cardiac cycles in each view were obtained at end-expiratory breath-hold and stored for offline analysis. Conventional spectral Doppler parameters were recorded, such as early (E) and late (A) diastolic transmitral flow velocities, isovolumic relaxation time, and septal and lateral mitral annular early diastolic (E′), late diastolic (A′), and systolic (S′) tissue velocities.
All image analysis was performed offline using EchoPAC version 1.13 (GE Healthcare, Little Chalfont, United Kingdom) by two experienced operators (A.R. and J.S.). LVEF was calculated using the Simpson biplane method. Speckle-tracking strain analysis was performed after manually adjusting the automatic detection of the epicardial and endocardial borders in each view and including all the apical myocardium (thus using an 18-segment model as opposed to the classical 17-segment one ). The software provided segmental and global peak longitudinal, circumferential, and radial strain, from which GLS and basal, midcavity, and apical circumferential and radial strain were calculated. LV rotation of the basal and apical levels (short-axis views) was calculated as the average angular displacement between the diastolic and systolic phases of the radial lines connecting the LV center to a specific epicardial point ( Figure 1 ). Rotation rates were defined as peak velocities of the respective clockwise (basal) or anticlockwise (apical) rotations as viewed from the apex. LV twist (in degrees) was defined as the net difference between apical anticlockwise rotation and basal clockwise rotation. Torsion was defined as twist normalized to the base-to-apex LV length measured at end-diastole (in degrees per centimeter). Twist rate was calculated as the peak velocity of twist (in degrees per scond). Early systolic “propping” twist (EST) was measured as the difference between the initial apical and basal rotation. This early motion occurs during isovolumic contraction and is directed clockwise at the apex and counterclockwise at the base (thus opposite to the normal direction of movement to follow during systole, like an athlete throwing a discus who first turns to his right before spinning leftward to throw); this early motion is thought to have a role in “propping” the myocardium for effective systolic twist. Similarly, diastolic apical and basal reverse rotation was measured as the peak clockwise and counterclockwise diastolic reverse rotations of the LV short-axis cross-sections, respectively. Untwist rate was determined from the rotational rate curve as the peak of the first negative deflection following aortic valve closure, as previously described. The times to peak rotation, reverse rotation, peak twist, peak EST, and peak untwist were measured from the peak of the R wave. To correct for any differences in heart rate, for all temporal analyses, the R-R interval was normalized to 100%. The interobserver reproducibility for speckle-tracking LV mechanics was 5 ± 2% (LV twist and GLS), 6 ± 2% (EST), and 3 ± 1% (untwist), each performed in n = 50 studies (20 with acute takotsubo syndrome, 20 at 4-month follow-up, and 10 control subjects by J.S. and A.R.). The intraobserver variability was 4 ± 2% (LV twist and GLS), performed in all subjects included in this study (J.S.).
A typical LV strain curve in a normal heart is shown in Figures 2 A and 2D.
Cardiac Magnetic Resonance
A 3-T Achieva scanner (Philips, Best, The Netherlands) was used for CMR imaging. The protocol included cine imaging, early and late gadolinium enhancement (Gadovist 0.1 mmol/kg; Bayer Schering Pharma, Berlin, Germany) with swap of the phase-encoding direction and pre- and postcontrast modified Look-Locker imaging T1 mapping. Data for T1 mapping were acquired with 3(3)3(3)5 and 5(3)3 schemes for native and postcontrast T1 measurement, respectively, the latter at exactly 15 min after contrast administration. All CMR images were each analyzed by a pair of two independent expert observers (T.A., C.S., A.R., C.J.N.) blinded to each other, to the other imaging modality, and for T1 mapping to the order of the scan. The CMR images were analyzed in CMR Tools (Cardiovascular Imaging Solutions, London, United Kingdom) for computation of LV volumes and mass. Each segment was scored for functional status (1 = normal, 2 = hypokinetic, 3 = akinetic, 4 = dyskinetic). T1 maps were generated using Philips RelaxMaps tool and quality-controlled with goodness-of-fit maps (χ 2 ). These were imported into Segment (Medviso, Lund, Sweden), in which T1 values were output for each of the 16 segments of the 17-segment model (omitting the true apex), and myocardial extracellular volume (ECV) was calculated as previously described. Our inter- and intraobserver variabilities for all T1 mapping analysis were 2.7 ± 1.5% and 1.5 ± 0.5%, respectively, and the interstudy variability for native T1 mapping in healthy volunteers was 2 ± 0.5%.
Data are expressed as mean ± SD for normally distributed data and otherwise as medians and interquartile ranges. Continuous variables were checked for normality of distribution using the Shapiro-Wilk test. Differences between acute presentation and follow-up data were examined with the paired Student’s t test or Wilcoxon matched-pairs signed rank test as appropriate. Patient data at follow-up were compared with those of control subjects using unpaired tests (unpaired Student’s t test or Mann-Whitney U test). Categorical variables were analyzed using χ 2 test or Fisher’s exact test as appropriate. To account for multiple comparisons, the P value chosen for statistical significance was Bonferroni-adjusted depending on the number of variables compared between groups, after which significance was chosen at P < .05. Data were analyzed using the GraphPad Prism version 5 (GraphPad Software, San Diego, CA) and SPSS version 24 (SPSS, Chicago, IL). Variability was calculated as mean ± SD of the ratios between differences and means of the two independently measured variables and expressed as a percentage.
Table 1 shows the patients’ and control subjects’ demographic and general clinical characteristics. The majority (79%) presented with chest pain, and 87% had stressful triggers. From the initial 52 acutely presenting patients, 44 returned for follow-up. Four patients died (two of cardiogenic shock during acute admission, one of arrhythmic cardiac arrest, and one of subsequent suicide), and four patients underwent automated implantable defibrillator or pacing device implantation (one with ventricular fibrillation and long-QT syndrome, one with asystole after direct-current cardioversion for paroxysmal atrial fibrillation, and two with complete heart block, one of them preceding the takotsubo acute event; these were excluded from imaging follow-up because pacing is likely to affect strain data and were unsuitable for CMR). All medications commenced for presumed acute myocardial infarction were stopped as soon as the diagnosis of takotsubo cardiomyopathy was established (i.e., after coronary angiography and CMR provided a clear-cut diagnosis). Our center’s takotsubo management policy for the duration of this enrollment was to not extrapolate on therapies used in other conditions with different pathophysiology (i.e., myocardial infarction) in the absence of any available evidence on the basis of “primum non nocere.” During their convalescence, patients continued only on the preexisting medical therapy for any comorbidities, except if LVEF was <50% at discharge, in which case angiotensin-converting enzyme inhibitors were continued empirically. The control subjects were of the same age and gender distribution, had similar cardiac comorbidities, and were on similar medications. Patients with takotsubo cardiomyopathy had a significantly higher heart rate, both acutely and at follow-up ( P = .001 for both), although their heart rates were within normal physiologic limits ( Tables 1 and 2 ).
|Demographics||Patients with takotsubo syndrome, acute presentation ( n = 52)||Patients with takotsubo syndrome, follow-up ( n = 44)||Control subjects ( n = 44)||P value, follow-up vs control subjects|
|Age (y)||66 ± 11||65 ± 10||67 ± 10||.418|
|Women||47 (92%)||40 (90%)||40 (90%)||≥ .999|
|BMI (kg/m 2 )||26 ± 4||26 ± 5||25 ± 2||.423|
|Hypertension||15 (28%)||10 (23%)||12 (28%)||.771|
|Diabetes||5 (9%)||4 (9%)||3 (8%)||.238|
|Thyroid disease||13 (26%)||11 (27%)||0||.004 ∗|
|Mental health disease||17 (34%)||14 (32%)||0||.024 ∗|
|Smoker/ex-smoker||2 (4%)/6 (11%)||2 (4%)/6 (11%)||1 (3%)/7 (18%)||.338|
|Alcohol||6 (11%)||5 (11%)||6 (13%)||.722|
|Aspirin||6 (11%)||5 (11%)||6 (13%)||≥ .999|
|β-blocker||5 (9%)||3 (7%)||4 (10%)||.238|
|ACE inhibitor/ARB||14 (27%)||10 (23%)||6 (13%)||.084|
|CCB||4 (8%)||3 (7%)||3 (8%)||≥ .999|
|Heart rate (beats/min)||80 ± 16||81 ± 16||67 ± 10||.001 ∗|
|Systolic BP (mm Hg)||127 ± 23||125 ± 23||129 ± 17||.113|
|Diastolic BP (mm Hg)||76 ± 16||76 ± 16||76 ± 10||.977|
|Chest pain||41 (79%)|
|Pulmonary edema||2 (4%)|
|ST-segment elevation||41 (79%)|
|Arrhythmia (VT/VF/CHB/asystole)||6 (11%)|
|Troponin I (ng/mL)|
|12 h||3.81 (1.2–8.8)|
|CRP (mg/L)||5.5 (<0.04–14.8)|
|BNP, presentation (pg/mL)||586 ± 720|
|BNP, follow-up (pg/mL)||48 ± 26|
|Coronary atheroma (<50%)||15 (28%)|
|LV angiography type|
|Patients with Takotsubo syndrome, acute presentation ( n = 52)||Patients with takotsubo syndrome, follow-up ( n = 44)||P value (acute vs follow-up, paired analysis)||Control subjects ( n = 44)||P value (follow-up vs control subjects, unpaired analysis)|
|LVEF (%)||45 ± 13||62 ± 9||<.0001 ∗||64 ± 6||.342|
|LV EDV (mL)||82 (71–102)||71 (61–86)||.006 ∗||67 ± 14||.138|
|LV ESV (mL)||44 (33–66)||28 (20–33)||.0001 ∗||23 (21–27)||.047|
|SV (mL)||39 ± 14||45 ± 12||.042 ∗||42 ± 9||.396|
|Moderate/severe MR||4/0||1/0||.643||0/0||≥ .999|
|LVOT Vmax (m/sec)||0.9 ± 0.2||0.9 ± 0.2||.357||1.0 ± 0.1||.361|
|AV Vmax (m/sec)||1.3 ± 0.3||1.4 ± 0.3||.130||1.4 ± 0.2||.572|
|E (m/sec)||0.7 ± 0.2||0.7 ± 0.1||.363||0.7 ± 0.1||.708|
|A (m/sec)||0.7 ± 0.1||0.7 ± 0.2||.312||0.7 ± 0.1||.959|
|E/A ratio||0.8 (0.6–1.1)||0.8 (0.7–1.2)||.435||0.9 (0.7–1.2)||.544|
|IVRT (msec)||121 ± 34||114 ± 32||.164||100 ± 21||.072|
|E′ (cm/sec)||5 ± 2||7 ± 2||.0002 ∗||8 ± 2||.179|
|S′ (cm/sec)||5 (4–6)||7 (5–8)||.002 ∗||7 (6–8)||.499|
|E/E′ ratio||11 (9–15)||9 (7–12)||.020 ∗||8 (7–11)||.310|
|TR Vmax (m/sec)||2.6 ± 0.6||2.4 ± 0.3||.170||2.3 ± 0.3||.265|
|Estimated RVSP (mm Hg)||35 (25–45)||30 ± 8||.057||27 ± 6||.149|
|LV EDV (mL)||132 (110–152)||126 (112–141)||.214||118 ± 16||.126|
|LV ESV (mL)||56 (46–83)||44 (36–54)||.000 ∗||41 ± 9||.308|
|LVEF (%)||53 (47–62)||64 (60–68)||.000 ∗||66 ± 4||.956|
|LVM (g)||139 (114–154)||118 (109–133)||.001 ∗||106 ± 21||.070|
|LV EDVi (mL/m 2 )||76 (65–85)||70 (63–83)||.206||69 ± 7||.938|
|LV ESVi (mL/m 2 )||34 (24–47)||26 (20–32)||.000 ∗||24 ± 5||.187|
|LVMi (g/m 2 )||80 ± 15||68 (57–78)||.001 ∗||61 ± 10||.061|
LV Strain and Twist at Presentation versus Follow-Up
As shown in Table 2 , there were marked abnormalities in most echocardiographic variables measured at acute presentation. Global indices of systolic and diastolic function such as LVEF and volumes as well as tissue Doppler indices of systolic (S′) and diastolic (E′, E/E′) function were abnormal upon presentation and improved significantly at follow-up ( P < .05 for all) ( Table 2 ). There were significant abnormalities at presentation compared with follow-up in LV strain ( Table 3 ): GLS, midcavity radial and circumferential strain, and apical radial and circumferential strain ( P < .05 all). Figure 3 shows a typical example of changes in apical circumferential strain.
|Strain and twist variables||Patients with takotsubo syndrome, acute presentation ( n = 52)||Patients with takotsubo syndrome, follow-up ( n = 44)||P value (acute vs follow-up, paired analysis)||Control subjects ( n = 44)||P value (follow-up vs control subjects, unpaired analysis)|
|GLS (%)||−10 ± 3||−16 ± 2||<.0001 ∗||−19 ± 1||<.0001 ∗|
|Radial strain (%)|
|Base||26 (16–38)||34 ± 14||.405||40 ± 13||.123|
|Mid||25 (16–38)||36 ± 17||.043 ∗||45 ± 11||.037 ∗|
|Apex||15 (6–23)||25 (16–40)||.006 ∗||22 (14–32)||.386|
|Circumferential strain (%)|
|Base||−12 ± 4||−13 ± 3||.665||−15 ± 4||.155|
|Mid||−11 (−14 to −7)||−14 ± 3||.014||−16 ± 4||.203|
|Apex||−11 ± 6||−17 ± 5||.0002 ∗||−22 ± 4||.0001 ∗|
|Apical rotation (°)||5.9 ± 5.8||7.2 (4.6 to 10.8)||.029 ∗||10.1 ± 6.5||.192|
|Apical rotational rate (°/sec)||40 ± 40||47 (37 to 80)||.057||50 (43 to 74)||.690|
|Time to peak apical rotation (msec)||340 ± 94||416 ± 96||.002 ∗||395 ± 62||.339|
|Basal rotation (°)||−5.0 ± 5.5||−5.7 ± 3.4||.798||−6.7 ± 6.0||.452|
|Basal rotational rate (°/sec)||−56 ± 27||−48 ± 37||.141||−68 ± 32||.023 ∗|
|Time to peak basal rotation (msec)||331 ± 95||404 ± 84||.014 ∗||384 ± 96||.216|
|LV twist (°)||11 ± 7||12 ± 6||.298||22 ± 4||<.0001 ∗|
|Time to peak LV twist (msec)||343 ± 80||404 ± 72||.001 ∗||384 ± 47||.244|
|LV torsion (°/cm)||1.4 ± 0.9||1.7 ± 0.9||.243||3.1 ± 0.8||<.0001 ∗|
|LV twist rate (°/sec)||80 ± 38||80 ± 35||.747||115 ± 23||<.0001 ∗|
|EST (°)||−0.7 (−1.8 to 0.0)||−1.5 ± 2.3||.255||−2.1 ± 1.3||.252|
|Missing EST||8 (25.8%)||1 (3.8%)||.070||0||≥ .999|
|Time to peak EST (msec)||62 (38 to 102)||75 ± 34||.868||53 ± 26||.011 ∗|
|Untwist rate (°/sec)||−76 ± 33||−77 ± 34||.824||−118 ± 27||<.0001 ∗|
|Time to peak untwist rate (msec)||505 ± 104||545 ± 76||.182||484 ± 79||.006 ∗|