Cardiovascular magnetic resonance assessment of acute cardiovascular effects of voluntary apnoea in elite divers

Inclusion criteria were experience in apnoea diving with a minimum breath-hold time of 270 s, a minimum age of 18 years, an unremarkable history of cardiac and lung disease, and the absence of long-term medication. Exclusion criteria were any contraindication to CMR or any known heart or lung disease. Participants were required not to drink caffeine-containing drinks and were instructed not to eat at least 8 h before the examination.

All study subjects received an information sheet 14 days prior to the study. Informed consent was obtained from all participants prior to study inclusion. Participants were questioned about training protocols and diving experience.

The study protocol comprised two apnoea sessions of individual maximum breath-hold combined with CMR measurements (one “functional cardiac session” and one “flow session”). Participants were asked to perform their usual pre-apnoea routines (yoga and breathing exercises). Fifteen minutes before CMR the participants were required to stop with their individual exercises and to breathe normally. A maximum of three deep inspirations prior to the final breath-hold was allowed. Hyperventilation was not allowed. Apnoea was performed as long as the individual subjects were able to withstand the breathing reflex.

This “individual” approach close to personal best breath-hold amplifies the redistribution effects and supposedly exhibits the maximum effect on the cardiovascular system. Additionally, a brain MRI was performed at least 30 minutes after the first of the two breath-hold sessions to detect acute brain ischemia. Cardiac biomarkers and catecholamines were evaluated to detect cardiac damage (for time points see Fig. 1).

All CMR studies were performed during voluntary breath-hold in maximal inspiration and in supine position using a 1.5 Tesla whole body scanner (Ingenia, Philips Healthcare, Best, The Netherlands). The protocol for the “functional cardiac session” consisted of retrospectively gated balanced steady state free precision cine imaging with 30 cardiac phases per slice. To assess functional changes under apnoea, three short axis (apical, midventricular, basal) cines as well as a vertical long axis cine  were acquired repeatedly over the course of apnoea. T2-mapping – indicative of myocardial oedema - using a gradient-spin-echo technique was performed in the same three slices in short axis orientation as the steady state free precision cine images prior and immediately after maximal apnoea [12].

MRI of the brain was performed using a 3 Tesla MRI scanner (Ingenia, Philips Healthcare) using a dedicated head coil. The protocol comprised a transverse T2-weighted turbo spin echo (TSE), a transverse fluid attenuated inversion recovery (FLAIR), a transverse T2*, a sagittal T1-weighted 3D gradient echo, and a transverse as well as a coronal diffusion weighted imaging (DWI) sequence.

The second CMR session was performed at least 4 h after the first “functional” cardiac session. This session was focused on flow measurements in the ascending and descending aorta, the pulmonary trunk, and both common carotid arteries using common 2D phase contrast imaging.

Two radiologists blinded to the study protocol independently evaluated all images. The first (apnoea_early) and last completed imaging data set (apnoea_late) of each maximum apnoea were compared. Global cardiac function at resting conditions (left ventricular end-diastolic volume (LVEDV), end-systolic volume (LVESV), stroke volume (LVSV), and ejection fraction (LVEF) were determined from the short axis (ViewForum, Philips Healthcare) and normalized to the body surface area (BSA) using Mosteller’s formula [13]. LVEDV and LVESV were quantified manually by tracing the endocardial borders in all short axis slices. For functional parameters under apnoea LV volumes were determined using the modified Simpson rule. For the assessment of regional cardiac function, fractional shortening (FS) was assessed as described previously [14]. In short, the endocardial distance from the free LV lateral wall to the septal wall was measured in end-diastole (EDD) and end-systole (ESD) in apical, midventricular and basal slices in short axis orientation. FS was then calculated as followed: \( FS=\frac{EDD- ESD}{EDD}\times 100 \). For flow quantification, phase contrast imaging was analysed using dedicated software (ViewForum, Philips Healthcare). Borders of the ascending and the descending aorta, the pulmonary trunk and both common carotid arteries (CCA) were manually traced in the magnitude images and the region of interest was automatically copied to phase images with manual correction performed when deemed necessary. Maximum velocity (V_max), mean velocity (V_mean), mean flow (Q_mean) and absolute stroke volume (SV) was assessed, respectively. T2 relaxation times were extracted from the T2 maps that were generated by using dedicated software (Intellispace Portal 9.0, Philips Healthcare). A circular region of interest (ROI) was then manually placed in the septal and lateral LV wall and averaged.

Cerebral MRI was analysed by the same two radiologists for focal diffusion restrictions as a sign of acute cerebral ischemia, cerebral micro-bleedings, and incidental findings. Peripheral oxygen saturation (SpO₂) and heart rate (HR) were measured continuously during both CMR sessions using a CMR-compatible device (Expression MR400, Invivo, Gainesville, Florida, USA).

Myocardial oxygen demand was estimated using the modified pressure work index as previously described (PWI mod = modified pressure work load index, P_systolic = systolic blood pressure, P_diastolic = diastolic blood pressure, HR = heart rate, CO = cardiac output, BSA = body surface area) [15]:

Urine was collected for baseline measurements 30 min before the first apnoea session started and 4 h thereafter, but before starting the second CMR session. Catecholamine levels, N-terminal pro-hormone of brain natriuretic peptide (NT pro-BNP), brain natriuretic peptide (BNP) and high sensitive troponin (hs-cT) were analysed from venous blood samples taken before, immediately after, 30 min, and 4 h after the first apnoea. All results (urine and blood samples) therefore reflect the effect of the first single breath-hold, but not those of repetitive apnoeal stages.

NT-pro BNP measurements were performed immediately after blood collection under routine conditions with the LOCI™-based NT-proBNP assay for Dimension™ VISTA 1500 (Siemens Healthcare Diagnostics, Eschborn, Germany). For BNP, hs-cT, and catecholamine analysis, aliquots were stored at − 80 °C. BNP and hs-cT were measured using commercially available, specific immunoassays (BNP and STA High Sensitive Troponin-I assay for ARCHITECT™, both Abbott Diagnostics, Wiesbaden, Germany). Plasma catecholamine levels were analyzed using a catecholamine reagent kit (Chromsystems Instruments & Chemicals GmbH; ord. no. 5000; Graefelfing, Germany) with a HPL Chromatography (Waters Corporation, Milford, Massachusetts, USA). Urine catecholamines were analysed by Bio Rad HPLC Agilent 1100 Series (Agilent Technologies, Waldbronn, Germany).

In all participants, LVEDV, LVESV, and LVSV increased from onset to the end of apnoea (122.9 ± 24.2 ml vs. 176.9 ± 26.4 ml, p < 0.001; 47.3 ± 16.8 ml vs. 75.9 ± 16.3 ml, p < 0.001; 75.6 ± 16.9 ml vs. 101.0 ± 22.9 ml, p = 0.003) (Fig. 2a), while LVEF (Fig. 2b) decreased from 61.8 ± 9.4% to 56.8 ± 8.2% (p = 0.04). Despite a significant decrease of HR (75 ± 23 bpm vs. 61 ± 12 bpm; p = 0.028) (Fig. 2c), LV-CO remained unchanged (5.5 ± 1.6 l/min vs. 6.1 ± 1.7 l/min, p = 0.88) (Fig. 2d). A representative image series of gradual LV enlargement is shown in Fig. 3. FS decreased over the course of apnoea from 33.0 ± 6.0% to 23.8 ± 4.4% (p < 0.001), with the decrease in the apical slice (41.5 ± 7.4% vs. 27.9 ± 7.2%, p < 0.001) and the midventricular slice (30.7 ± 9.0% vs. 21.6 ± 4.4%, p < 0.001), whereas only a non-significant trend was observed in the basal slice (26.7 ± 7.1% vs. 22.6 ± 5.5%, p = 0.065) (see also Table 2).

Changes in HR (∆HR) over the course of apnoea (Fig. 4) showed a significant negative correlation with the changes of LVSV (∆LVSV) (Spearman’s rank correlation analysis; correlation coefficient − 0.64, p = 0.008). Additionally, ∆HR over the course of apnoea had a significant negative correlation with the change of LVEDV (∆LVEDV) (Spearman’s rank correlation analysis; − 0.59; p = 0.016). This was less prominent when ∆LVEDV was normalized to BSA (Spearman’s rank correlation analysis; − 0.55; p = 0.028). In contrast, there was only a weak correlation between ∆HR and an increase in ∆LVESV (− 0,319; p = 0.23).

While SV, V_max, V_mean, and Q_mean increased significantly in the ascending aorta and the pulmonary trunk during apnoea, no changes were observed in the descending aorta, indicating a preferential blood flow distribution to the heart and the brain. In addition, both CCAs showed a significant increase in SV (see Fig. 5), V_mean and Q_mean over the course of apnoea. All flow measurements are summarised in Table 3. A relevant shunt was neither observed at rest nor under apnoea (Qp/Qs: 1.06 ± 0.25 vs. 1.06 ± 0.19; p = 0.97).

We found a decrease in HR (76 ± 23 bpm vs. 61 ± 12 bpm) and an increase in LVSV (75.6 ± 16.9 ml vs. 95.1 ± 32.6 ml) in this study. Using a previously reported increase of systolic and diastolic blood pressure from 135 ± 13 mmHg to 185 ± 25 mmHg [16], the estimated myocardial oxygen demand using the modified pressure work index [15] increases from 8.51 ml/min/100 g to 9.48 ml/min/100 g (increase of 11%) during apnoea.

Serum catecholamine levels (see Fig. 6a, b) showed a fast, increase immediately after apnoea onset (epinephrine from 67.5 ± 23.9 pg/ml to 173.8 ± 113.2 pg/ml (p < 0.001); norepinephrine from 590 ± 197 pg/ml to 2063 ± 1703 pg/ml (p < 0.001)). Serum catecholamine levels returned to baseline values as early as 30 min post apnoea (epinephrine: 50.9 ± 26.6 pg/ml; norepinephrine: 474.5 ± 138.1 pg/ml). Catecholamine levels derived from urine samples still showed a slight, but significant increase 4 h after apnoea compared to baseline conditions (epinephrine from 6.1 ± 2.0 pg/ml to 11.3 ± 6.1 pg/ml, p = 0.003; norepinephrine from 25.0 ± 20.1 pg/ml to 42.3 ± 22.6 pg/ml, p = 0.011).

NT pro-BNP increased slightly from baseline levels 45.9 ± 40.3 pg/ml to 49.3 ± 43.3 pg/ml immediately after apnoea (p = 0.011) and to 53.8 ± 49.4 pg/ml (p = 0.037) 4 h after breath-hold (see Fig. 6c). BNP could not be quantified in 7 out of 17 subjects due to values lower than the detection limit (< 10 pg/ml). Overall, there were no significant changes of BNP serum levels at any time point.

Hs-cT increased from baseline until 4 h after apnoea (2.2 ± 1.1 pg/ml vs. 3.1 ± 1.7 pg/ml, p = 0.026) (Fig. 6d). Compared to baseline levels, this resulted in a mean relative Hs-cT increase of 56%, but was still far from any pathological range.

DWI revealed neither acute nor sub-acute signs of cerebral ischemia. In one participant, a clinically irrelevant singular micro-bleeding formation located in the brain stem was observed. In two participants, unilateral fluid collections of the mastoid were observed and reported. No further incidental findings were observed.

In the present study we were able to demonstrate a significant LV dilatation along with an increased LVSV, which is in line with a previous study [17], where increased EDD and ESD, an increase in SV and CO and a reduction in contractile function after an apnoea time of 3.7 ± 0.3 min was reported. In contrast to our results, neither bradycardia nor increased calculated systemic vascular resistance were observed [17], although both effects are part of the accepted concept of the diving response [8]. In a more recent study by Batinic et al., cardiac parameters (i.e. HR, LV volumes, LVEF, LVCO) taken at two time points of apnoea (minute 1 and minute 3) were compared [18]. These investigators found a significant increase in LVEDV and CO (112 ± 15 ml to 125 ± 15 ml; 5.4 ± 1.9 l/min to 6.0 ± 1.2 l/min), which was similar to our results (123 ± 24 ml to 177 ± 26 ml; 5.5 ± 1.6 l/min to 6.1 ± 1.7 l/min). In contrast to the previous results from Pingitore et al. [17] and the results of the present study, no changes in SV were observed (69 ± 12 ml to 69 ± 8 ml), while HR increased from 80 ± 15 bmp to 87 ± 16 bmp during apnoea [18].

Since the mammalian diving response to maximal voluntary apnoea considerably varies depending on the examined individual and the study setup, the at first apparently contradictory results of the three studies might be explained by the breath-hold duration [5, 9, 19]. In contrast to previous studies focusing on physiological changes during apnoea, the breath-hold time in the present study was considerably longer (297 ± 99 s vs. 234 ± 66 s; 199 ± 11 s; 210 ± 70 s) [5, 9, 19]. However, even though individual responses may vary, it is known that physiological changes are most notable at the end of apnoea [9, 20, 21]. In this context it is important to mention that the previous studies [17, 18] used predefined time points for data collection, which will not necessarily coincide with the individual maximum breath-hold duration of each athlete. We have therefore decided to use a minimal breath-hold duration of 270 s to eliminate the possible shortcomings of a too short apnoea duration in the previous studies [5, 9, 19‐21]. Therefore, one can speculate that the shorter breath-hold durations registered in both previous studies [10, 17] are not suitable to push all compensatory mechanisms to their limits, and that a predefined time point might lead to undersampling. This is further supported by the fact that SpO₂ decreased more profoundly in our study compared to the study from Pingitore et al. (from 99 ± 1% to 74 ± 14% vs. 97 ± 0.2% to 84 ± 2%) [17].

We found a relative increase in LVSV of 30 ± 48% during apnoea, but a decrease in FS and LVEF. FS depends on inter-ventricular dimensions and is affected by ventricular filling. Ejection fraction, in contrast, is a relatively load independent surrogate parameter for cardiovascular performance. In general, the efficiency of myocardial performance is determined by preload, afterload and contractility [22, 23]. An increase in afterload will therefore result in decreased efficiency of myocardial performance. In case of prolonged breath-hold the peripheral chemoreflex regulation, the elevated sympathetic nerve activity and the increase in norepinephrine will lead to peripheral vasoconstriction and hypertension [5, 24] and subsequently to bradycardia via the baroreflex [25]. In accordance with this established physiological pathway, we observed a significant increase in norepinephrine levels to above the upper cut-off limit of > 420 pg/ml and a decrease in HR at the end of apnoea. Therefore, the HR decrease and the concommitant increase of both ventricles may be seen as an indirect visualization of the aforementioned baroreflex (Fig. 4).

NT-proBNP was elevated early after maximal apnoea. Although some authors describe BNP as an “emergency” cardiac hormone against ventricular overload [26], the observed elevations of pro-BNP were only minor and far from pathological levels. Nevertheless, in absence of other triggers even this small increase may be regarded as an indicator for LV wall stress. Although an increase in hs-cT was found in this study, the normal T2 relaxation times directly after apnoea may indicate that the increased hs-cT may be more attributable to the LV dilatation and not to acute and persistent myocardial damage. This may further be supported by the fact that elevated cardiac troponin (cT) levels are also commonly found in patients with dilated cardiomyopathy [27].

In addition, myocardial perfusion and oxygen consumption is dependant on various parameters. At the end of apnoea, HR decreases while SV and systolic and diastolic blood pressure increase. These physiological changes translate into an increase of estimated oxygen demand in our study from 8.5 ml/min/100 g to 9.5 ml/min/100 g (i.e. only by 11%). However, this increase in demand may be assumed to be outweighed by a theoretical increase of approximately 40% of coronary perfusion due to increase of the diastolic blood pressure. It is of note that these theoretical considerations are based on healthy subjects without any coronary morbidities.

The human diving response (i.e. bradycardia, peripheral vasoconstriction, increased blood pressure) helps to preserve O₂ in case of apnoea [28]. These protective mechanisms against hypoxia are triggered by apnoea per se and are augmented by face immersion [29]. The constriction of intramuscular and dermal vessels results in an increased total peripheral resistance and thus in an increased blood pressure [9, 30]. Due to peripheral vasoconstriction and reduced blood flow, the remaining circulating blood flow is redistributed to more hypoxia sensitive organs such as the brain [19, 25]. In the present study, only minimal blood flow changes were seen in the descending aorta while blood flow in the ascending aorta and the carotid arteries massively increased, indicating that even the gastrointestinal tract is excluded from blood flow redistribution in the case of hypoxia. This perfusion preference of the cerebrum emphasises the efficiency of the body’s compensatory mechanism to avoid hypoxic damage of the brain. Accordingly, cerebral MRI showed no case of acute ischemia-induced brain injury, indicating the effectiveness of the compensatory mechanisms, even in case of breath-holds longer than 8 min (Table 1, subject 12).

Prolonged apnoea is not exclusively seen in breath-hold divers, also patients with obstructive sleep apnoea (OSA) show compensatory mechanisms to avoid brain damage [31]. Patients with OSA show an increase in cerebral blood flow [32, 33], elevated sympathetic activity [34], elevated arterial blood pressure [35], and an increase in norepinephrine levels [31]. Interestingly, LV and right ventricular (RV) afterload are increased and cardiac arrhythmia is commonly seen [36]. OSA is independently associated with coronary artery disease, atherosclerosis, hypertension, stroke, endothelial function and myocardial infarction [37, 38]. A main problem in understanding the underlying pathophysiology stems from the lack of an adequate clinical model to simulate OSA [39]. So far, hypoxic gas mixtures have been used to mimic hypoxia in humans [40], but because of the resulting hyperventilation, these models are more representative for high altitude environments than for OSA. In addition, the transmissibility of animal models is also limited. Apnoea divers are mostly free of comorbidities, and our study shows that even a short episode of hypoxia affects the cardiovascular system. Therefore, voluntary extended breath-hold might be taken as a clinical relevant model to simulate short term changes due to hypoxia [41], although the exposure levels to hypoxemia differ significantly [41]. In this context it should be noted that this study was performed with trained athletes, and that a transfer of these findings to patients with cardiovascular diseases and obstructive sleep apnoea should be done with caution.

Patent foramen ovale has been demonstrated to have a higher prevalence in patients with obstructive sleep aponoea compared to healthy controls, and is suspected to inrease nocturnal oxygen desaturation in these patients [42] and to enhance other pathologic conditions associated with OSA [43]. In both scuba and apnoea divers knowledge about the implications of a patent foramen ovale regarding incidence and severity of decompression sickness is scarce [44], especially because it is unkown if recurrent decompression sickness is a result of a patent foramen ovale, the inabiliy to adopt a more conservative diving style, or both [45]. In the present study, no relevant changes of Qp/Qs (the stroke volume in the ascending aorta relative to the stroke volume in the pulmonary trunk) and thus no indication for a cardiac shunt, was found.

Cardiac dysrhythmia or irregular heartbeats (mainly premature ventricular excitations) were observed in 14 of 17 divers at the end of apnoea and during the early recovery phase (example shown in Additional file 1: Figure S1). It is tempting to speculate that the massive LV and RV dilatation triggers cardiac depolarization and repolarization. However, ECG quality was limited in this study and did not allow for a comprehensive analysis.

Measurment accuracy (CMR and SpO₂) might be limited at the end of apnoea due to e.g. motion artefacts (CMR), peripheral vasoconstriction (SpO₂), and other technical restrictions. Blood pressure data is not available as invasive blood pressure measurement was not performed due to ethical considerations and automatic non-invasive blood pressure measurement failed due to the high and dynamic changes in blood pressure during apnoea. Due to the chosen CMR imaging protocol with only limited coverage of the RV, neither volumetric nor functional RV data are available. Future studies should also focus on effects of hypxoxia on pulmonary vasoconstriction and their effects on the RV function.