Syncope is a sudden loss of consciousness caused by transient cerebral global hypoperfusion with immediate and spontaneous recovery [
Head-up tilt testing (HUTT) is a well-established tool in the diagnosis of VVS. It allows to diagnose vasovagal syncope, usually benign in its clinical presentation [
The assessment of haemodynamic response in HUTT is usually based on
Novel diagnostic tools, such as impedance cardiography (ICG) and plethysmography, enable continuous monitoring of other cardiovascular parameters, such as stroke volume (SV), cardiac output (CO), and systemic vascular resistance (SVR) [
Quantitative complexity theory (QCT) stems from so-called complexity science [
Eighty-one healthy volunteers (74 men and 7 women; mean age: 37.8 ± 4.7 years) were included in this retrospective analysis. The data were collected as a part of project no. 126/IWSZ/2007, funded by the Polish Ministry of National Defence and realized in Department of Cardiology and Internal Diseases, Military Institute of Medicine between January 2012 and October 2014. The project was approved by the appropriate ethics committee (no 11/WIM/2009) and performed in accordance with the ethical standards set out in the 1964 Declaration of Helsinki and its later amendments. The subjects aged 25–45 years, active soldiers, and without any chronic diseases were enrolled to this project. All participants provided written informed consent. Anonymity was ensured in all cases.
The subjects underwent HUTT, according to a modified version of the Italian Protocol [
Beat-to-beat haemodynamic cardiovascular response to tilting was evaluated by ICG, a modern, noninvasive method of haemodynamic monitoring. A Niccomo™ device (Medis, Ilmenau, Germany) integrated with a Tensoscreen™ module (Medis), dedicated to beat-to-beat BP assessment, was used. The final analysis included the following haemodynamic parameters: diastolic, systolic, and mean BP; pulse pressure; HR; pre-ejection period; left ventricular ejection time; stroke volume; cardiac output; Heather index; systemic vascular resistance; total artery compliance; and thoracic fluid content (described in detail in Table
In our approach [
Examples of scatter plots and corresponding images. Images were obtained by subdividing the area of a scatter plot into pixels. The intensity of each pixel is proportional to the number of data points falling into it.
This approach avoids the drawbacks of conventional linear techniques, which can, for example, miss significant correlations (see Figure
Examples of linear and generalised correlations.
In the present analysis, a moving window of 100 samples of beat-to-beat haemodynamic cardiovascular parameters was applied. The data sampling frequency corresponded to HR frequency. This means that for an HR of 60 bpm, a window of 100 spans 100 s. The size of moving window was based on previous empiric observations (data not published).
The obtained results were analysed statistically with Statistica 12.0 software (StatSoft Inc., Tulsa, OK, USA). The following time points were analysed: 300, 240, 210, 180, 150, 120, 90, 60, and 30 s before termination of the HUTT (HUTT_end). The distribution and normality of the data were assessed via visual inspection and the Kolmogorov–Smirnov test. Continuous variables were presented as means ± standard deviation (SD), and categorical variables were presented as absolute and relative frequencies (percentages). For comparative analysis (between subgroups with and without positive HUTT), the Student
The study participants were characterised by a mean BP of 117.9 ± 12.0/74.6 ± 7.7 mmHg and mean HR of 58.8 ± 8.9 bpm. They were free from chronic diseases and reported good physical fitness and regular physical training. The mean body mass index was 25.9 ± 2.6 kg/m2 (79 were nonobese). Only two subjects were current smokers. In 54 participants (66.7%), HUTT was positive and resulted in VVS (pre)syncope. No significant intergroup differences were noted (Table
Intergroup comparison of baseline characteristics.
Variable | Syncope_[yes] | Syncope_[no] | |
---|---|---|---|
Female/male | 51 (94%)/3 (6%) | 23 (85%)/4 (15%) | 0.162 |
Age (years) | 38.1 ± 4.6 | 37.1 ± 4.8 | 0.368 |
Office HR (bpm) | 58 ± 9 | 60 ± 9 | 0.541 |
Office SBP (mm Hg) | 117 ± 10 | 119 ± 16 | 0.573 |
Office DBP (mm Hg) | 75 ± 7 | 74 ± 9 | 0.738 |
BMI (kg/m2) | 26 ± 3 | 26 ± 3 | 0.470 |
Creatinine (mg/dL) | 0.94 ± 0.9 | 0.91 ± 0.13 | 0.139 |
Hemoglobin (g/dL) | 14.5 ± 0.9 | 14.8 ± 1.1 | 0.279 |
Data presented as mean ± standard deviation (SD) and
The values of complexity were already higher in fainting subjects 300 s before HUTT_end, with a significant upward trend starting 150 s before (pre)syncope (Table
Intergroup comparison of complexity, mean arterial pressure (MAP), and heart rate (HR) in the last 300 s before the termination of HUTT.
Time to HUTT_end (seconds) | Variable | Syncope_[yes] | Syncope_[no] |
---|---|---|---|
300 | Complexity (bits) | 625.6 ± 848.2 | 481.3 ± 462.3 |
HR (bpm) | 77.6 ± 13.0 | 88.9 ± 16.2 | |
MAP (mm Hg) | 103.2 ± 16.5 | 99.1 ± 16.8 | |
270 | Complexity (bits) | 726.5 ± 1077.4 | 501.9 ± 509.5 |
HR (bpm) | 79.4 ± 12.8 | 88.8 ± 16.9 | |
MAP (mm Hg) | 102.8 ± 18.2 | 99.0 ± 17.6 | |
240 | Complexity (bits) | ||
HR (bpm) | 79.5 ± 13.4 | 88.8 ± 13.8 | |
MAP (mm Hg) | 102.6 ± 17.9 | 98.0 ± 17.4 | |
210 | Complexity (bits) | 746.4 ± 1142.9 | 449.1 ± 386.7 |
HR (bpm) | 82.0 ± 14.7 | 86.2 ± 17.0 | |
MAP (mm Hg) | 103.8 ± 19.1 | 98.9 ± 17.4 | |
180 | Complexity (bits) | ||
HR (bpm) | 86.4 ± 17.5 | 86.3 ± 14.7 | |
MAP (mm Hg) | 103.8 ± 19.0 | 98.9 ± 17.1 | |
150 | Complexity (bits) | ||
HR (bpm) | 85.9 ± 15.3 | 86.5 ± 13.6 | |
MAP (mm Hg) | 103.2 ± 19.1 | 97.6 ± 18.2 | |
120 | Complexity (bits) | ||
HR (bpm) | 88.9 ± 17.5 | 87.9 ± 14.7 | |
MAP (mm Hg) | 104.1 ± 18.7 | 98.1 ± 18.4 | |
90 | Complexity (bits) | ||
HR (bpm) | 90.8 ± 19.5 | 86.4 ± 15.3 | |
MAP (mm Hg) | 103.1 ± 18.8 | 97.1 ± 17.5 | |
60 | Complexity (bits) | ||
HR (bpm) | 91.1 ± 21.6 | 85.1 ± 14.2 | |
MAP (mm Hg) | 101.2 ± 19.9 | 97.4 ± 17.3 | |
30 | Complexity (bits) | ||
HR (bpm) | 83.3 ± 21.6 | 85.1 ± 14.2 | |
MAP (mm Hg) | 93.9 ± 20.7 | 96.4 ± 19.8 |
Data presented as mean ± standard deviation (SD). HR: heart rate, HUTT: head-up tilt test, MAP: mean blood pressure. Statistically significant intergroup difference (syncope_[yes] vs syncope_[no]) is marked as
Intergroup comparison of complexity, mean arterial pressure (MAP), and heart rate (HR) values in the last 300 s before the termination of HUTT (
The results of the ROC analysis for complexity, MAP, and HR are presented in Table
Prognostic value of complexity, mean arterial pressure (MAP), and heart rate (HR) in the last 300 seconds before the termination of HUTT.
Time to HUTT_end (seconds) | Variable | AUC (95% CI) |
---|---|---|
300 | Complexity (bits) | 0.527 (0.393–0.661) |
HR (bpm) | 0.713 (0.591–0.834)0.026 | |
MAP (mm Hg) | 0.585 (0.452–0.718) | |
270 | Complexity (bits) | 0.576 (0.440–0.712) |
HR (bpm) | 0.632 (494–0.770) | |
MAP (mm Hg) | 0.564 (0.429–0.698) | |
240 | Complexity (bits) | 0.683 (0.558–0.808) |
HR (bpm) | 0.713 (0.594–0.833) | |
MAP (mm Hg) | 0.574 (0.441–0.707) | |
210 | Complexity (bits) | 0.566 (0.431–0.700) |
HR (bpm) | 0.583 (0.447–0.718) | |
MAP (mm Hg) | 0.577 (0.448–0.706) | |
180 | Complexity (bits) | 0.673 (0.553)-0.793) |
HR (bpm) | 0.510 (0.377–0.644) | |
MAP (mm Hg) | 0.573 (0.442–0.704) | |
150 | Complexity (bits) | 0.660 (0.539–0.781) |
HR (bpm) | 0.520 (0.386–0.655) | |
MAP (mm Hg) | 0.578 (0.445–0.711)0.050 | |
120 | Complexity (bits) | 0.747 (0.634–0.860) |
HR (bpm) | 0.504 (0.369–0.638)0.034 | |
MAP (mm Hg) | 0.578 (0.444–0.712)0.106 | |
90 | Complexity (bits) | 0.736 (0.625–0,847) |
HR (bpm) | 0.535 (0.405–0.665)0.007 | |
MAP (mm Hg) | 0.572 (0.437–0.707) | |
60 | Complexity (bits) | 0.802 (0.707–0.897) |
HR (bpm) | 0.590 (0.465–0.716)0.015 | |
MAP (mm Hg) | 0.546 (0.414–0.677)0.003 | |
30 | Complexity (bits) | 0.772 (0.660–0.883) |
HR (bpm) | 0.525 (0.398–0.652)0.014 | |
MAP (mm Hg) | 0.550 (0.417–0.683)0.030 |
Comparison of ROC curves of complexity, mean arterial pressure (MAP) and heart rate (HR) for the chosen time points: 180 s (a), 120 s (b), and 60 s (c) before the termination of HUTT.
The optimal cutoffs of complexity (column A) within last 300 seconds before the termination of HUTT, corresponding sensitivity and specificity (column B), specificity in term sensitivity over 80% (column C), and sensitivity in term specificity over 80% (column D).
time to HUTT_end (seconds) | Column A -optimal cutoff | Column B -sensitivity/specificity for optimal cutoff | Column C -cutoff (specificity) when sensitivity over 80% | Column D -cutoff (sensitivity) when specificity over 80% | |
---|---|---|---|---|---|
300 | Complexity (bits) | 199.3 | 83%/26% | 193.8 (26%) | 689.2 (20%) |
270 | Complexity (bits) | 231.7 | 78%/44% | 194.3 (26%) | 715.7 (30%) |
240 | Complexity (bits) | 246.9 | 76%/56% | 208.8 (48%) | 611.5 (51%) |
210 | Complexity (bits) | 231.7 | 78%/44% | 188.5 (26%) | 713.7 (26%) |
180 | Complexity (bits) | 598.0 | 50%/85% | 230.4 (41%) | 591.1 (50%) |
150 | Complexity (bits) | 319.1 | 61%/70% | 184.2 (33%) | 470.1 (48%) |
120 | Complexity (bits) | 411.5 | 63%/78% | 240.6 (52%) | 523.7 (52%) |
90 | Complexity (bits) | 456.5 | 65%/74% | 130.8 (37%) | 586.1 (50%) |
60 | Complexity (bits) | 385.6 | 82%/67% | 385.6 (67%) | 584.2 (59%) |
30 | Complexity (bits) | 545.7 | 69%/74% | 371.7 (59%) | 753.0 (54%) |
HUTT, head-up tilt test.
HR only presented a significantly higher AUC than complexity did at 300 s before HUTT_end, but the closer to the HUTT termination, the prognostic power of HR expired. There was also no intergroup difference for HR in time points from 210 s to 30 s before HUTT_end. The prognostic value of MAP was poor at all the analysed time points (max. 0.585).
Our results suggest that complexity analysis based on QCT may be valuable in predicting the HUTT outcome, presenting additional value to HR and BP monitoring. Clinical application of QCT could avoid the necessity to impose a full syncopal event and make HUTT a less traumatic experience for the patient. In our study, HR and MAP were revealed to be inapplicable in predicting syncope, which can be explained by different haemodynamic patterns of vasovagal reaction that cannot be anticipated before HUTT (vasovagal, cardiodepressive, mixed).
Many subjects have little or no warning or prodromal symptoms before fainting [
Several earlier studies investigated haemodynamic predictors of syncope in the late phase of HUTT. Mereu et al. [
There were also some attempts to predict the HUTT result based on resting supine measurements and the early phase of HUTT. Schang et al. [
Predicting syncope appears to be a difficult task because the cardiovascular response to upright posture depends on complex mechanisms. Faes et al. [
The application of QCT, an advanced mathematical model integrating several parameters into one marker, allowed for the prediction of the HUTT outcome with clinically acceptable power and timing.
The previous reports described above basically used post hoc, sophisticated, analytical methods. Their advantage is that most of them provide abundant data and explain, at least to some extent, pathophysiological background of syncope. The importance of detailed investigating physiological mechanisms of cardiovascular interactions accurately illustrates the studies such as that reported by Javorka et al. [
The relevant advantage of QCT is the emerging possibility of real-time analysis and delivering patient monitoring software for mobile devices, such as smartwatches or tablets. QCT merges multiple streams of haemodynamic data into one parameter and provide quantitative and holistic information on the cardiovascular reaction to orthostatic challenge. However, it is also a potential disadvantage. Such an approach might be perceived as a “black-box,” missing a detailed insight into the mechanism of vasovagal reaction. Although the contribution of haemodynamic parameters might be derived from complexity profiles (Figure
This study had some limitations. First, the present research was a retrospective analysis of a single-centre study with all the accompanying limitations inherent to such a design. Second, HUTT was performed in healthy volunteers without a spontaneous syncope history. The high occurrence of HUTT-induced syncope was probably related to accumulation of factors predisposing to orthostatic intolerance and specific for this study group, especially young age and higher than normal exercise training load. Therefore, our results are not conclusive of the diagnostic value of HUTT in general population. Moreover, the contribution of our approach to the clinical practice should be derived from further research in target populations, such as patients with a history of syncopal events. A comparison of QTC to other diagnostic methods discussed above while designing further studies could be very valuable.
The findings of intergroup differences in HR at 300 and 240 s before HUTT, as though relatively high AUC at 300 s before HUTT for HR, should also be commented. The causal relationship of this observation with the result of HUTT seems to be unlikely. Firstly, the results are inconsistent with the expectation that as syncope approaches, the prognostic power of the predictor increases. Secondly, the intergroup comparison of trends (Figure
We are also aware that the description of QCT in this article is limited, but the technology is proprietary by one of the coauthors (JM) and certain details cannot be disclosed.
Complexity has been shown to be a sensitive marker of cardiovascular haemodynamic response to orthostatic stress and proved its superiority over HR and BP in predicting HUTT outcomes. The predictive performance (sensitivity >80% and specificity >50%) of complexity even 2 min before syncope seems to be clinically acceptable. Beat-to-beat complexity analysis may be used to terminate HUTT before triggering a symptomatic vasovagal reflex with a high probability of correct diagnosis. The present results encourage validating the complexity method in other clinical settings and in prospectively designed trials.
The data used to support the findings of this study are available from the corresponding author upon request.
The data were collected as a part of the project (no. 126/IWSZ/2007) funded by the Polish Ministry of National Defense.
Jacek Marczyk has propriety rights to QCT. Bartosz Wolszczak is an employee of Ontonix s.r.l.
The authors would like to thank all medical stuff from Department of Cardiology and Internal Diseases involved in performing hemodynamic measurements.
Supplementary Materials include the list of haemodynamic parameters measured by impedance cardiography (Table S1) and examples of complexity profiles showing the breakdown of total system complexity into its components in terms of the percentage of contribution of each component (Figure S1).