Post-exercise contractility, diastolic function, and pressure: Operator-independent sensor-based intelligent monitoring for heart failure telemedicine

We enrolled 172 consecutive subjects referred for semi-supine exercise-stress echocardiography, mean age 55 ± 18 years. Subjects comprised 22 controls, 52 patients with CAD, 10 with dilated cardiomyopathy (DCM), 13 with hypertension, 18 with chronic obstructive pulmonary disease (COPD), 14 with valvular disease, 13 with corrected congenital heart disease and 30 with previous non-diagnostic exercise testing. Patient characteristics are summarized in Table 1. Coronary artery disease was defined by the presence of angiographically assessed coronary stenosis (with quantitatively assessed diameter reduction > 50% in at least one major coronary vessel) or previous myocardial infarction; patients with dilated cardiomyopathy had LV end-diastolic volume > 140 ml/m², LV end-systolic volume > 70 ml/m²; 22 non-competitive athletes were the controls. The local Ethical Committee approved the study protocol. All patients gave their written informed consent before entering the study. All patients met the following inclusion criteria: 1) referred to stress echo for clinically-driven testing, 2) acoustic window of acceptable quality, and 3) willingness to enter the study, 3) recent (within 1 year) coronary angiography for all patients with stable chest pain syndrome or dilated cardiomyopathy. Exclusion criteria were: 1) unstable angina or recent myocardial infarction and 2) technically poor baseline echocardiographic examination. From the initial population of 178 patients, 6 were excluded due to a poor acoustic window (n = 4), or refusal to give written informed consent (n = 2).

Graded semi-supine bicycle exercise echo was performed, starting at an initial workload of 25 watts lasting for 2 min; thereafter the workload was increased stepwise by 25 watts at 2-min intervals. A 12-lead electrocardiogram and blood pressure measurement were performed at baseline and every minute thereafter [10]. Two-dimensional echocardiographic monitoring was performed throughout and up to 5 min after the end of stress.

Standard echocardiographic data, echo-derived hemodynamic parameters and systemic pressure were measured at peak stress and at 1, 3 and 5 min post-exercise [11, 12].

Regional wall motion analysis was evaluated at baseline and at peak stress with a semiquantitative assessment of a wall motion score index (WMSI), with the 17-segment model of the left ventricle, each segment ranging from 1 = normal/hyperkinetic to 4 = dyskinetic, according to the recommendations of the European and of the American Society of Echocardiography [13, 14]. WMSI was derived by dividing the sum of individual segment scores by the number of interpretable segments [10, 13]. Test positivity was defined as the occurrence of at least one of the following conditions: 1) new dyssynergy in a region with normal rest function (i.e. normokinesia becoming hypokinesia, akinesia or dyskinesia) in at least two adjacent segments [15, 16]; 2) worsening of a resting dissynergy.

The diagnostic end-points were the development of obvious echocardiography positivity. The test was also stopped, in the absence of diagnostic endpoints, for one of the following reasons for performing a submaximal, non-diagnostic test: intolerable symptoms, limiting asymptomatic side effects consisting of (a) hypertension (systolic blood pressure > 220 mmHg; diastolic blood pressure > 120 mmHg) (b) relative or absolute hypotension (> 30 mmHg fall in blood pressure), (c) supraventricular arrhythmias: supraventricular tachycardia or atrial fibrillation, (d) ventricular arrhythmias: ventricular tachycardia; frequent, polymorphous premature ventricular beats [10].

One nurse recorded blood pressures of each patient at rest and during the study. The blood pressure recording was made using a sphygmomanometer and the diaphragm of a standard stethoscope [17]. Systolic and diastolic blood pressure was measured in the right arm. During exercise test and recovery, blood pressure was recorded with the patient lying in a left-rotated semi-supine position and gripping the left support with their left hand. Patients were told to let their right hand go limp when blood pressure was measured. Post-exercise hypotension was defined as at least 5 mmHg of diastolic blood pressure reduction vs the same exercise heart rate values [18].

By protocol, the first 52 consecutive enrolled subjects underwent both complete echocardiographic and sensor-based hemodynamic assessment at rest, peak exercise, and at the first, third, and fifth minute post-exercise (Table 2).

All the 172 enrolled subjects, scheduled for exercise stress, had standard ECG, pressure and WMSI evaluation and complete (both stress and recovery) sensor evaluation. The transcutaneous force sensor was based on a linear accelerometer of the LIS3 family (STMicroelectronics [Geneva, Switzerland]). The device includes in a single package a MEMS sensor that measures capacitance variation in response to movement or inclination and a factory trimmed interface chip that converts the capacitance variations into analog signals proportional to the motion. The device has a full scale of ± 2·g (g = 9.8 m/s²) with a resolution of 0.0005·g. We housed the device in a small case (Fig. 1) which was positioned in the mid-sternal precordial region and was fastened by a solid gel ECG electrode. The acceleration signal is acquired along with an ECG signal and transmitted to a laptop PC by wireless connection [19]. The data are analyzed using software developed in Matlab (The MathWorks, Inc Natick, Massachusetts, USA). A peak detector algorithm synchronized with the ECG scans the first 150 ms following the R-wave to record first heart sound force vibrations and the 100 ms following the T-wave to record second heart sound force vibrations (Fig. 2). These values are then filtered by a median filter, which averages the beat-to beat variations in the signal and removes outliers caused by movement artifacts. Contractility quantification through the first heart sound (FHS) amplitude recording and systemic arterial pressure changes through the second heart sound (S2) amplitude recording have been previously demonstrated [4, 6]. Apart from the first and the second heart sound amplitude (related to the isovolumic contraction force and to the isovolumic relaxation force) this recording system was utilized to quantify both cardiological systole and diastole duration [5]. According to the physiological background, cardiological systole was demarcated by the interval between the first and the second heart sounds, lasting from the first heart sound to the closure of the aortic valve. The remainder of the cardiac cycle was automatically recorded as cardiological diastole [7] (Fig. 2). Information obtained from the ECG (QRS detector) and the heart sound vibration (HSV peak detector): heart rate, first heart sound and second heart sound are analyzed by algorithms in order to derive the systolic force-frequency relation, diastolic time-frequency relation and pressure-frequency relation [4‐6]. The signals are displayed in real-time and processed by special-purpose software (Fig. 3). All the parameters were acquired at baseline, during stress and recovery; mobile mean was used to assess values. Values during exercise and recovery were computed as absolute value, and as percent changes vs baseline.

The systolic force-frequency relation, the diastolic time-frequency relation and the second heart sound derived systemic pressure were recorded at both stress and recovery in the 172 recruited subjects to exploit sensor-based dichotomy patterns in the leap to non-invasive intelligent remote heart monitoring.

SPSS 11 for Windows was used for statistical analysis. The statistical analyses included descriptive statistics (frequency and percentage of categorical variables and mean and standard deviation of continuous variables). Since we observed large differences in baseline sensor-derived systolic force and S2 vibration amplitude, % changes were assessed during stress and recovery for interpatient and intergroup comparisons. Diastolic times were measured as absolute values simultaneously with systolic/diastolic time ratio. Chi-square test was used for comparisons between echo and sensor based contractility overshoot assessment (yes/no) and reduced pressure (yes/no) in the post- exercise phase. Changes in continuous variables during recovery were compared by analysis of variance for repeated measures. When this test was significant, individual comparisons of end-exercise value (recovery time = 0) and values during 1, 3 and 5 min of recovery were made by Duncan's multiple-range test. Since all the sensor-based parameters are physiologically heart-rate dependent, comparisons of recovery minutes 1, 3 and 5 values were made with sensor-recorded values at the same heart rates during exercise. A p-value < 0.05 was accepted as statistically significant.

Exercise time was 11 ± 5 min in the 22 controls and 9 ± 3 min in the 150 patients. WMSI was 1 at rest = peak stress in the 22 controls and 1.17 ± 0.35 at rest, 1.18 ± 0.37 peak stress in the 150 patients. Four patients had stress-induced ischemia (WMSI rest = 1.36 ± 0.35, peak = 1.77 ± 0.31).

The first 52 consecutive enrolled subjects (7 controls) underwent both echocardiographic (Table 2) and sensor hemodynamic assessment (Fig. 4) at rest, peak exercise, and the first, third, and fifth minute of the post-exercise phase. At recovery minute 1 we observed a decrease in LV end-systolic volume, increased stroke volume index, and a decreased systemic diastolic and systolic pressure. Since contractility changes are measurable both with echo (SP/ESV index) and with the sensor (FHS vibrations amplitude), comparisons were made. Heart rate dropped rapidly at recovery, but at each recovery step contractility was higher than exercise values in 42 subjects with echo and in 40 with sensor (Chi-square p < 0.01). Since systemic pressure changes are measured by protocol and derivable by the sensor (S2 vibrations amplitude), comparisons were made (Fig. 4, lower panels); 41 (80%) of the subjects had post-exercise hypotension, and 36 (69%) second heart sound undershoot (Chi square p < 0.01). Diastolic time length was easily measured only by the sensor: at each recovery heart rate diastole was longer than during exercise (Fig. 4).

Sensor data for contractility, diastolic time-frequency relation, and second heart sound were obtained in all the 172 enrolled subjects (feasibility = 100%) and are displayed in Fig. 5. Contractility comparisons between different groups of patients based on the incoming disease are reported in Fig. 6.

During exercise the force frequency relation was upsloping in controls and blunted in patients. Post-exercise contractility overshoot (defined as a relative increase in recovery contractility of more than 10% with respect to the exercise value) was more frequent in patients than controls (27% vs 8%, p < 0.05) (Fig. 7). The overshoot phenomenon was found in 2 controls and 45 patients (p < 0.05), and more frequently in DCM (50%), congenital (69%), hypertensive (39%), previous MI (36%) vs CHD (21%) or COPD (22%) or diagnostic (17%) patients, chi-square p = 0.003.

From rest to peak exercise, the mean systolic time was shortened by 25 ± 11%. The diastolic time decreased more markedly during exercise (by 53 ± 13% at peak stress) [5]. At 100 bpm heart rate during exercise, 20 patients (and at peak stress, 63 subjects) showed a reversal of the systolic/diastolic ratio, with the duration of systole longer than that of diastole (Fig. 8). The diastolic time increased abruptly during the first minute of recovery, with an overshoot phenomenon. At each recovery heart beat frequency, the diastolic time was higher than the diastolic time recorded during exercise, all p < 0.05 vs exercise. At 100 bpm heart rate during recovery, only three patients still showed reversal of the systolic/diastolic ratio, with the duration of systole longer than diastole (p < 0.05 vs exercise). In the post-exercise period, the sensor-based quantification of the diastolic time-frequency relation (diastolic time at each value of decreasing heart rate) showed that diastole lengthened in the post-exercise phase in both controls and patients: at recovery 100 bpm heart rate, + 39 ± 22 msec lengthening in controls vs + 33 ± 21 msec lengthening in patients, p = ns. However, at the individual level a broad spectrum was found in patients with a -30 msec to + 81 msec range.

A significant correlation was found between post-exercise hypotension and recovery second heart sound amplitude (S2) undershoot: 138 subjects had normal post-exercise hypotension; 83% of the subjects with post-exercise hypotension had S2 undershoot in the recovery, while 84% of the 34 subjects without post-exercise hypotension had stable rate-S2 curve at recovery (Fig. 9).

At force-frequency analysis, with both computed elastance (SP/ESV index) and operator-independent sensor (FHS vibrations amplitude) most patients showed a contractility increase at recovery. Standard systemic pressure measures (obtained by sphygmomanometer) were blunted in the post-exercise phase, and a general decrease in the sensor-based derived systemic pressure were found [6]. Furthermore, during recovery as during stress, the cutaneous operator-independent force sensor described systolic and diastolic duration in real time. Simultaneous calculation of stroke volume with echo and diastolic time with force sensor allowed us to monitor the diastolic filling rate [5].

During exercise the normal force-frequency relation is upsloping, while a flat-biphasic force-frequency relation is abnormal [20‐24]. A post-exercise sensor-based contractility overshoot was more frequent in patients vs controls, and frequently associated with an abnormal blunted force-frequency relation during exercise. Several investigators using different methods [12, 25‐27] have reported an overshoot of cardiac function during recovery from maximal exercise in patients with cardiac disease. Koike et al. [26] have shown a marked rebound of stroke volume immediately after cessation of upright exercise in patients with previous myocardial infarction. Tanabe at al. [27] found an overshoot of cardiac output at 1 min of recovery in patients with severe CHF, along with poor cardiac output response to exercise. They found that not only O₂ uptake but also cardiac output fell much more slowly after maximal exercise as CHF worsened. Although insufficient afterload reduction during exercise in CHF contributes to the impaired stroke volume response to exercise, systemic vascular resistance at 1 min of recovery significantly decreased from that at peak exercise in patients with CHF who showed overshoot of cardiac output during recovery [28‐31]. The marked increase in stroke volume during early recovery in patients with overshoot appears to result from both an immediate afterload reduction and a relatively slow decrease in cardiac sympathetic stimulation during recovery [32, 33]. Knowledge of the post-exercise overshoot also had prognostic value since patients with a moderate exercise intolerance and a normal recovery period had a better prognosis than a patient with a post-exercise overshoot [34].

We found that at each recovery heart beat frequency, the diastolic time was higher than the diastolic time recorded during exercise, in both controls and patients. Why does diastolic time increase in the post- exercise phase? The total cardiac cycle duration is algebraically dependent on the heart rate [= 60,000 msec/heart rate] [35]. However at each heartbeat frequency the fixed total cardiac cycle time can be differently divided between systole and diastole [5]. The diastolic time fraction is determined by factors that modulate systolic duration through modulation of myocyte contraction [36‐40]. At recovery, improved myocardial contractility and reduction in systemic vascular resistance significantly shortened left ventricular ejection time, with a proportionate increase in diastolic time fraction. Stress-induced "systolic-diastolic mismatch" can be easily quantified by a disproportionate decrease in diastolic time fraction, and is associated with several cardiac diseases, possibly expanding the spectrum of information obtainable during stress [5]. The post-exercise diastolic time fraction indicates the duration of absence of compression of intramural vessels during a heartbeat and has a dominant role in the subendocardial layer – whose perfusion is mainly diastolic, whereas the perfusion in the subepicardial layer is also systolic [36, 38]. The lengthening of cardiological diastole is much more pronounced than lengthening of cardiological systole, and the former is much more effective for subendocardial perfusion, even in the absence of coronary artery disease. This indicates the relevance of monitoring both exercise and recovery diastolic time in the critically diseased heart [37, 39].

A significant correlation was found between post-exercise hypotension and recovery S2 undershoot (Fig. 9). In this investigation, blood pressure (systolic, diastolic and mean) correlated closely with S2 amplitude during both stress and recovery [6]. This could be explained by the fact that amplitude is primarily determined by one factor, the force of valve closure [41, 42]. In the selected patients of our study, a significant correlation was found between post-exercise hypotension and recovery second heart sound lower amplitude, to confirm the sensor's ability to mirror the diastolic pressure trend. Post-exercise hypotension has been demonstrated in both hypertensive and healthy subjects [43], and it has been attributed to a decrease in cardiac output and/or systemic vascular resistance [44‐47]. Acute exercise may serve as a non-pharmacological aid in the treatment of hypertension. S2 amplitude monitoring could be a method for assessing the efficacy of acute post-exercise blood pressure reduction.

Congestive heart failure (HF) is a serious public health problem due to its prevalence, high mortality, high morbidity, and the expense of ongoing therapy [2].

Several strategies to control fluid volume status are used in the practice. Clinic visits for assessment of filling pressure by physical examination, multiple types of non-invasive measurements, and repeated cardiac catheterization may be employed. There is considerable cost and inconvenience for the patient associated with these strategies and, more importantly, these methods represent pressure and volume status only as one discrete point in time without the perturbance of daily activities or stress. A system of frequent monitoring could alert clinicians to early signs and symptoms of decompensation, providing the opportunity for intervention before patients become severely ill and require hospitalization [1].