Exercise cardiac MRI unmasks right ventricular dysfunction in acute hypoxia and chronic pulmonary arterial hypertension

X Shareen Jaijee, Marina Quinlan, Pawel Tokarczuk, Matthew Clemence, Luke S.G.E. Howard, J. Simon R. Gibbs,* and Declan P. O’Regan* MRC London Institute of Medical Sciences, Imperial College London, London, United Kingdom; Philips Healthcare, Philips Centre, Guildford, United Kingdom; Department of Cardiology, National Pulmonary Hypertension Service, Imperial College Healthcare NHS Trust, London, United Kingdom; and National Heart and Lung Institute, Imperial College London, London, United Kingdom


INTRODUCTION
Pulmonary arterial hypertension (PAH) is a disease characterized by adverse remodeling of the peripheral pulmonary arteries leading to elevated pulmonary arterial pressure (PAP) and progressive right ventricular (RV) dysfunction (5,17). In the early stages of PAH, the RV remains coupled to its afterload and maintains efficient energy transfer by a homeometric adaptive increase in wall thickness and contractility. With cumulative exposure to increasing afterload, heterometric adaptation occurs with progressive RV dilatation to preserve stroke volume (SV) until the more advanced stages of disease, when wall stress rises and ventriculoarterial uncoupling occurs with reduced cardiac output (26). It is known that RV function is a strong predictor of mortality, outperforming pulmonary vascular resistance (PVR) (18,33,35), but there is currently no method to determine which RV phenotype is destined to fail despite medical therapy. The change in systolic function on exercise or pharmacological stress is emerging as a more promising indicator of coupling efficiency (13), and the response of the pressureloaded RV to exercise could be more relevant than resting hemodynamics for the followup and management of patients with PAH (11).
RV reserve can be assessed noninvasively during exercise using real-time imaging to provide a surrogate marker of stress-induced cardiopulmonary uncoupling (16,21). Understanding the consequences of increased vascular load on RV contractile reserve may allow earlier risk stratification and evaluation of treatment response (38). In the present study, we used real-time exercise imaging to determine the effect of elevated afterload on RV contractile reserve during normobaric hypoxia (which has been associated with transient acute pulmonary hypertension) and in chronic PAH before the RV becomes dilated or impaired. We hypothesized that the RV has limited capacity to recruit contractile function when exposed to elevated afterload despite compensated function at rest.

Study Participants
We prospectively enrolled patients undergoing routine treatment in the Pulmonary Hypertension Service at Imperial College Healthcare National Health Service Trust and normal subjects from August 2014 to February 2015. Ethical approval was given by the Health Research Authority, United Kingdom, and informed written consent was obtained from all participants. Patients were diagnosed with group 1 PAH according to joint European Society of Cardiology and European Respiratory Society guidelines and received standard treatment (8). Patients were excluded if they had a contraindication to cardiac magnetic resonance (CMR) imaging, exercise testing, were unstable, or had a resting RV ejection fraction (RVEF) of Ͻ45%.

Study Design
Patients with PAH underwent CMR at rest and exercise while breathing room air, whereas normal subjects underwent an exercise protocol while breathing room air and then during normobaric hypoxia.

Cardiopulmonary Exercise Test
The cardiopulmonary exercise test (CPET) comprised a rampprotocol cycle test on an electronically braked bicycle ergometer (VIASprint 200, BD Worldwide). Normal subjects were encouraged to exercise until exhaustion, after which 3 min of rest and recovery data were recorded. Each participant breathed through a calibrated mass flow sensor with expired gas sampled on a breath-by-breath basis.

Exercise Protocol
All participants performed a supine exercise protocol while lying in the bore of the magnet using an magnetic resonance-compatible cycle ergometer (Lode, Groningen, The Netherlands). CMR images were acquired with subjects breathing room air in the resting state and then at an exercise intensity threshold set at 40% of the watts achieved during upright CPET with a cadence of 60 -70 revolutions/min. This protocol accounts for interindividual variation in exercise capacity as well as the difference in efficiency between upright (CPET) and supine (CMR) exercise (6), providing accurate assessment of cardiac output compared with the direct Fick method (21). In normal subjects, the protocol was repeated while subjects breathed 12% O 2 (ensuring resting saturations had dropped and plateaued). The total duration of hypoxia was~35 min.
CMR imaging. CMR was performed on a 1.5-T Philips Achieva system (Best, The Netherlands), and a standard clinical protocol for structural, functional, and phase-contrast breath-hold imaging at rest was followed according to published international guidelines (20). Real-time imaging at both rest and exercise were then acquired using a previously validated ungated high-temporal resolution sequence during free breathing (Philips Healthcare Clinical Science, Surrey, UK) (21) using the following parameters: field of view, 305 ϫ 305 mm; repetition time/echo time, 2.5/1.26 ms; flip angle, 50°; acquired voxel size, 2.73 ϫ 2.73 ϫ 10 mm; section thickness, 10 mm with no gap; reconstructed voxel size, 1.19 ϫ 1.19 ϫ 10 mm; number of sections, 14; slice scan order, base to apex; dynamic scan time, 66 s with 50 dynamics per slice; and temporal resolution, 74 ms. Free breathing, real-time main pulmonary artery and aortic phase-contrast images were acquired using the following parameters: field of view, 300 ϫ 300; repetition time/echo time, 11/3.7 ms; flip angle, 20°; acquired voxel size, 3.0 ϫ 3.0 ϫ 10 mm; section thickness, 10 mm; reconstructed voxel size, 1.17 ϫ 1.17 ϫ 10 mm; dynamic scan time, 4.5 s with 100 dynamics; and temporal resolution, 44 ms.
Ventricular volumes and function analysis. A standard clinical protocol for assessing biventricular function, volumes, mass, and flow was followed according to published international guidelines (20). Real-time imaging data were analyzed using a specially adapted proprietary software (cvi42, Circle Cardiovascular Imaging, Calgary, AB, Canada) and assessed during the expiratory phase of respiration. Volumes were indexed to body surface area calculated using the Mosteller formula. Indexed volumetric data were left ventricular (LV) and RV end-diastolic volumes (LVEDVi and RVEDVi, respectively), LV end-systolic volume and RV end-systolic volume (LVESVi and RVESVi, respectively), LV and RV stroke volumes (LVSVi and RVSVi, respectively), and LV ejection fraction (LVEF) and RVEF. Cardiac index (CI) was derived as follows: [LVSVi ϫ heart rate (HR)]. RV contractile reserve was defined as RVEF with exercise Ϫ RVEF at rest (13). Real-time septal curvature measurements were assessed using previously described techniques (29 Values are expressed as medians (interquartile ranges) and were compared with a Mann-Whitney test. Max load CPET, maximum watts achieved during cardiopulmonary exercise test; load CMR, level at which the subject was exercised during the cardiac magnetic resonance imaging; PAP, pulmonary arterial pressure.

Reproducibility
Ten normal subjects returned for repeat exercise CMR imaging during normoxia within 3 mo of the initial scan for interstudy reproducibility. Inter-and intraobserver variability was assessed in 15 randomly selected data sets of both volume and flow at rest and exercise.

Statistical Analysis
Data were analyzed using SPSS version 22 (IBM, Armonk, NY). Descriptive data for continuous variables are presented as means Ϯ SD or as medians (interquartile ranges). Test-retest reliability was assessed using an intraclass correlation coefficient (ICC) with a two-way random model for absolute agreement. Comparisons between groups for continuous variables were performed using unpaired two-sample t-tests or the Mann-Whitney test and comparison between categorical variables with Fisher's exact test. Pearson product-moment correlation was used to assess the linear relationship between variables. The effect of O 2 level and exercise on cardiac parameters was assessed with two-way repeated-measures ANOVA, and the effect of normoxic exercise in patients and normal subjects was compared with mixed ANOVA. The F-test remains a valid statistical procedure under non-normality in a variety of conditions (3). Interactions were first evaluated with simple main effects analysis. Predictors of RV reserve were assessed using stepwise multiple linear regression. Bonferroni correction was made for multiple comparisons. P values of Ͻ0.05 were considered significant.

RESULTS
Thirty-eight normal subjects were investigated (age 35.5 Ϯ 9.3 yr, 16 men and 12 women). Thirty-four subjects completed the protocol; one subject withdrew at the onset of the resting hypoxia and three subjects were excluded because they did not complete the hypoxic exercise protocol because of fatigue, and these normal subjects were excluded from the analysis. Fourteen patients with PAH were investigated (39.1 Ϯ 9.4 yr, 4 men and 10 women) a mean of 3.9 Ϯ 3.3 yr after the initial diagnosis. Thirteen patients had idiopathic PAH based on right heart catherization criteria, and one patient had hereditary PAH (heterozygous bone morphogenetic protein receptor type 2 mutation). CPET was performed 1.7 Ϯ 2.2 mo before CMR, and the RHC data collected were at diagnosis. Baseline characteristics at diagnosis are shown in Table 1; treatments are shown in Table 2. Intra-and interobserver ICCs for resting RVEF were 0.71 and 0.85, respectively, and for exercise RVEF were 0.625 and 0.744. The interstudy ICC for exercise RVEF was 0.603. The effect of each exercise condition on RVEF and LVEDVi in patients and control subjects is shown in Fig. 1.

Biventricular Response to Normoxic and Hypoxic Exercise in Healthy Participants
Biventricular parameters during normoxia and hypoxia at rest and exercise are shown in Table 3. There was a significant effect of exercise on all cardiac parameters except RVEDVi (P between Ͻ0.0001 and 0.031). During normoxic exercise, normal subjects increased their CI by 2.5 times the resting value, at individualized workloads, with significant increases of HR, SV index (SVi), biventricular EF, and RVSVi/RVESVi. An increase in LVSVi was associated with increased end-diastolic volumes and decreased end-systolic volumes. In contrast, the increase in RVSVi was associated with decreased RVESVi but stable RVEDVi (all significant differences P Ͻ 0.0001). In normal subjects, breathing a fraction of inspired O 2 of 12% reduced peripheral O 2 saturation to 84 Ϯ 4% at rest and 77 Ϯ 5% during exercise with the development of concomitant interventricular septal flattening (P Ͻ 0.0001), which was present in 85% of subjects (Fig. 2). During resting hypoxia, HR increased by 21% and CI by 16% (P Ͻ 0.0001). Left atrial size at rest was significantly reduced by hypoxia (24.1 Ϯ 5.4 to 21.0 Ϯ 9.3 cm 2 , P Ͻ 0.05), but right atrial size was unchanged (24.8 Ϯ 5.9 vs. 25.0 Ϯ 5.4 cm 2 , P ϭ 0.621). For most cardiac variables, the response to exercise depended on O 2 level. There was no difference between resting SVi in normoxia and hypoxia (P ϭ 0.76), but exercise SVi was significantly lower during hypoxia compared with normoxia (P Ͻ 0.001). The increase in RVEF on exercise was less during hypoxia (P ϭ 0.017), but the response of LVEF to exercise did not depend on O 2 level. Biventricular volumes decreased during resting hypoxia, and hypoxic exercise conditions accentuated the fall in RVEDVi and attenuated the rise of LVEDVi (P Ͻ 0.036 for all interactions).

Biventricular Response to Exercise in Patients with PAH
Biventricular parameters in patients compared with control subjects are shown in Table 4. There were no significant differences in resting indexed RV volumes and function, HR, and CI between patients and normal subjects. During exercise, patients increased their CI by 1.8 times the resting value with a significant rise in HR. In contrast to control subjects, there was no overall response of RVEF to exercise, but the positive response of LVEF was not significantly different in patients. The lack of increase of RVSVi on exercise was associated with a marked increase in RVESVi compared with control subjects (P Ͻ 0.0001) despite a modest increase in RVEDVi (P Ͻ 0.001). RVSVi/RVESVi also showed no augmentation with exercise. Similarly, patients were unable to increase LVSVi despite a decrease in LVESVi on exercise, with no increase in end-diastolic volume.

DISCUSSION
We investigated biventricular contractile reserve using realtime CMR during acute normobaric hypoxia in healthy adults and in patients with compensated PAH (RVEF Ͼ 45%). Noninvasive cardiac imaging during exercise unmasks reduced RV contractile reserve in healthy adults under hypoxic conditions and a marked depletion of reserve in patients with PAH under normoxic conditions despite preserved resting systolic function. Exercise imaging of the pressure-loaded right heart during exercise offers a means to stratify patients with normal resting function noninvasively.
A moderate increase in pulmonary vascular tone over the first 5 min of normobaric hypoxia has been observed in heathy adults at rest using Doppler echocardiography (34), and a similar response has been observed in animal models using invasive measurements (39). As the pulmonary vascular pressure-flow relationship is steep, there is a substantial afterload on the RV during exercise (26). Resting hypoxia induced a significant decrease in peripheral O 2 saturation in our normal subjects, which further decreased during hypoxic exercise. We observed that the increase of RVEF on exercise in normal subjects was attenuated by acute hypoxia and was associated with a fall in SVi despite sympathetic activation and homeometric adaptation. In contrast, the positive response of LVEF to exercise remained the same during hypoxic and normoxic exercise. Our data show that hypoxia causes mildly impaired contractile reserve selectively affecting the RV, although the RVSVi-to-RVESVi ratio suggested that coupling was maintained (37). Sildenafil may reverse the negative impact of hypoxia on RV contractility during exercise in some subjects (19), and taken with our findings, it is plausible that the RV limits exercise capacity during hypoxic vasoconstriction (26).
During exercise, an increase in mean PAP is partially offset by a decline of PVR in healthy adults, but in patients with PAH, PVR is fixed, placing even greater contractile demands on the RV (27). Furthermore, in the context of PAH, contractile reserve is dependent on multiple factors, such as ventricular contractility, myocardial fibrosis, and response to neurohormonal activation (13). Previous imaging studies in patients with PAH with dilated and impaired RVs at rest have shown   (14,23), but our data show that diminished contractile reserves are also a characteristic of patients with preserved resting RV systolic function and volumes, with up to half of patients exhibiting a fall in RVEF on submaximal exercise. In normal subjects, the increase in SVi on normoxic exercise was most likely determined predominantly by enhanced contractility (2RVESVi Ͼ Ͼ 2RVEDVi). In contrast, patients with PAH were not able to augment SVi, with RVEDVi and RVESVi both shifting by similar amounts to higher operating volumes suggestive of uncoupling of the RV and pulmonary arteries (15). While there are several mechanisms by which contractile reserve might be impaired in PAH, our data from hypoxic exercise indicates that even the healthy RV has limited capacity to recruit contractile function in the face of rising afterload. There is a strong relationship between ventriculararterial coupling and RV contractile reserve in animal models of chronic pressure overload (12), supporting the potential role of stress testing as a surrogate marker for exercise-induced uncoupling in PAH (12,16,32). Exercise-induced changes in RVEDVi and pressure-mediated septal curvature interfere with LV diastolic filling, but our data also suggest that an impaired response to exercise may contribute to reduced LV preload and consequent underfilling.
Clinical deterioration in PAH is preceded by alterations in resting RV volumes (36), and progressive loss of function is associated with a poor outcome, irrespective of any changes in PVR (1,18,33,35). Conventional assessment of baseline function is often a weak predictor of the transition from a compensated state to right heart failure, and hence detecting RV dysfunction using exercise MRI before it manifests as resting RV dilatation and dysfunction could unmask ventricular-vascular uncoupling (9,25,31,40). Using exercise imaging to risk-stratify patients by RV contractile reserve and to determine response to therapy is worthy of further investigation and would offer a noninvasive alternative to the hemodynamic stress tests, which are known to be strong predictors of prognosis in left heart failure, PAH, and chronic thromboembolic pulmonary hypertension (4,7,13). Previous studies have also implicated poor RV performance and higher exercise pulmonary vascular tone in the pathogenesis of ventilatory inefficiency (22), and we have shown that the minute ventilation-to CO 2 output ratio is independently associated with RV contractile reserve in PAH. The mechanism by which dynamic changes in PVR during exercise mediate inefficient ventilation is debated, but the minute ventilation-to CO 2 output ratio has been shown to be the single best indicator of increased risk of adverse events in heart failure (28) and is predictive of outcome at baseline in PAH (10). Impaired RV function may contribute to ventilatory inefficiency and exercise-related dyspnea even in well-compensated patients with PAH.

Limitations
Exercise CMR using ungated continuous imaging is considered the "gold standard" for ventricular volume measurement during high-intensity exercise and shows close agreement with invasive assessment, but its accuracy depends on adequate temporal and spatial resolution to avoid underestimating SV at high HRs (21,24). The best metric of contractile reserve is still an area of investigation, and although inotrope-induced change in ventricular elastance is the reference standard, in clinical practice contractile reserve is usually defined by a change in EF or SV during exercise or with dobutamine (13). Acute hypoxia leads to rapid pulmonary vasoconstriction and has been used in numerous studies of normal subjects, at rest and exercise, to investigate how the respiratory and cardiovascular systems respond to acute rises in pulmonary pressures (2,26,39). However, acute hypoxia not only directly affects the vascular tone of the pulmonary and systemic vessels but also causes sympathetic nervous activation leading to enhanced HR, cardiac output, and myocardial contraction (27). In hypoxia, septal curvature measurements indicated a significant rise in the transeptal pressure gradient, but we did not measure pulmonary hemodynamics during CMR.

Conclusions
Noninvasive cardiac imaging during submaximal exercise demonstrates that RV contractile reserve decreases during hypoxic vasoconstriction in healthy control subjects and that patients with chronic PAH show marked depletion of contractile reserves despite preserved EF at rest. Exercise imaging has the potential to unmask dysfunction in the RV before resting adaptations to PAH are evident.