Exercise Tolerance In Patients With Idiopathic Pulmonary Artery Hypertension: Insight Into Risk Thresholds And Prediction Capacity For 5-Year Mortality


 Background: Exercise tolerance is pivotal for risk-stratification in patients with idiopathic pulmonary artery hypertension (IPAH), yet optimal risk predictors and risk thresholds remain uncertain. This study aimed to investigate risk estimates of cardiopulmonary exercise testing (CPET) associated with 5-year mortality in patients with IPAH and explore their risk thresholds and prediction capacity.Methods: Consecutive patients with IPAH who underwent right heart catheterization and CPET were retrospectively enrolled and followed up for five years. Multivariable Cox proportional hazards models were used to determine independent prognostic factors for mortality. The risk trend and threshold for mortality were exhibited using restricted cubic splines. Survival rates were estimated by Kaplan-Meier analysis stratified by various CPET parameters.Results: Among 210 patients with IPAH (75.7% female), 37 (17.6%) died during a 34-month median follow-up. Three CPET variables were independently predictive of mortality in multivariable Cox regression analysis (all P<0.05), including oxygen uptake efficiency slope (OUES), peak oxygen pulse (VO2/HR), and peak oxygen consumption (VO2), in descending order of prediction power (𝑥2 = 37.39 > 35.96 > 35.57). The levels of OUES at 0.91, peak VO2/HR at 5.3 ml·min-1·beat-1, and peak VO2 at 12.2 ml·kg-1·min-1 were respectively identified as risk thresholds for mortality. Patients below these thresholds had significantly higher mortality risk (adjusted hazard ratio of OUES: 3.34; peak VO2/HR: 3.76; and peak VO2: 1.56) and lower survival rates (log-rank test, all P<0.01). The joint model (area under the curve [AUC] 0.838) of these CPET variables (AUC 0.724) and estimates in contemporary risk assessment tools (AUC 0.809) provided more excellent prediction capacity for 5-year mortality. Conclusions: Suboptimal exercise tolerance indicated by OUES, peak VO2/HR, and peak VO2 under certain thresholds posed a higher mortality risk in patients with IPAH, and their joint combination further improved the prediction capacity.


Introduction
Idiopathic pulmonary artery hypertension (IPAH) is a rare but devastating disease characterized by progressive remodeling of the small pulmonary arteries, leading to increased pulmonary vascular resistance, right heart failure (HF), and even death [1]. Despite advances in modern therapeutics for pulmonary artery hypertension (PAH), such as various combinations of drugs and modes of administration, the long-term prognosis remains unsatisfactory [2]. Exploring prognostic indicators of adverse outcomes is thereby warranted to facilitate a more accurate and urgent risk strati cation.
Cardiopulmonary exercise testing (CPET) permits the evaluation of exercise intolerance [3]. It has become a valuable tool that has the potential of non-invasively estimating disease severity, therapeutic response, and prognosis in patients with pulmonary hypertension (PH) [4][5][6][7][8][9][10]. Parameters that have been shown to embody these capabilities included peak oxygen consumption (VO 2 ), ventilation/carbon dioxide output slope (VE/VCO 2 slope), end-tidal partial pressure of carbon dioxide at anaerobic threshold (PETCO 2 @AT), and oxygen uptake e ciency slope (OUES). In recent years, there has been emerging data on the superiorities of OUES in predicting outcomes in patients with HF [11,12] and PAH [7,8] over other ventilatory e ciency parameters such as VE/VCO 2 slope. However, ndings into CPET variables predictive of poor outcomes have been inconsistent and contradictory [4][5][6][7][8][9][10]. Notably, most of these variables and cut-off values were based on expert opinion or receiver operator characteristic curves (ROC). The dynamic risk range and thresholds for these CPET metrics associated with adverse outcomes, and the additional role in PAH risk-strati cation according to current guidelines, were less explored.
Therefore, the purpose of this study was (1) to determine and compare the prognostic signi cance of risk parameters of CPET for predicting 5-year mortality, and (2) to explore their risk thresholds and whether they can further improve the prediction capacity compared to contemporary assessment tools for patients with IPAH.

Participants
Consecutive patients admitted to Fuwai Hospital from January 2015 to January 2020 with a newly diagnosed IPAH were retrospectively enrolled in this cohort. The diagnosis of IPAH was con rmed through right heart catheterization (RHC) according to the 2015 European Society of Cardiology (ESC)/European Respiratory Society (ERS) PH guidelines. Baseline clinical data including demographics, World Health Organization functional class (WHO-FC), 6-minute walk distance (6MWD), and key laboratory tests were collected upon admission. The study was approved by the Fuwai Hospital Ethics Committee (No. 2018-1100). Written informed consent was obtained from all enrolled participants.

Cardiopulmonary Exercise Testing
All recruited patients underwent CPET using the COSMED Quark CPET system before receiving speci c therapy. The test was conducted by an experienced medical staff blinded to patients' medical records, and the equipment was calibrated before the individual test. Oxygen therapy was ceased for at least 30 minutes before performing CPET. After a 3-minute resting and a subsequent 3-minute warm-up of unloaded pedaling, patients started to exercise on a cycle ergometer with electromagnetic brake at a progressively incremental work rate of 5-30 W/min according to the estimated exercise tolerance until reaching volitional exhaustion or symptoms limitation. Gas exchange parameters were measured by a metabolic cart on a breath-by-breath basis and averaged over 10-second intervals. Peak VO 2 was de ned as the highest obtained average oxygen consumption measured over 30 seconds in the last minute of exercise. Peak VO 2 /heart rate (HR) was calculated as peak VO 2 divided by peak heart rate. The ventilatory AT was detected by combining the V-slope method and ventilatory equivalents [13]. The VE/VCO 2 slope was identi ed as the slope of the linear regression relationship between minute ventilation (VE) and carbon dioxide production from resting to the peak exercise. The OUES represented the slope of the regression line between VO 2 (y axis, L·min -1 ) and the logarithmically transformed VE (x axis, L·min -1 ) across the entire exercise course. Heart rate was measured every 1 minute. Heart rate reserve was calculated as the difference between peak and resting heart rate. Blood pressure was recorded at 3minute intervals and when the patients expressed peak exercise.

Hemodynamic Studies
Hemodynamic assessment with RHC was performed at resting to facilitate the diagnosis of IPAH.
Parameters composed of mean pulmonary artery pressure (mPAP), pulmonary vascular resistance (PVR), pulmonary capillary wedge pressure (PCWP), right atrial pressure (RAP), mixed venous oxygen saturation, and cardiac output. Diagnostic criteria of IPAH include an mPAP ≥ 25 mmHg, and PVR > 3 Wood units at rest in the presence of a normal PCWP ≤ 15 mmHg when none of the de nite PH etiologies (e.g., connective tissue disease, congenital heart defect, heritable, drugs or toxins-induced, severe lung or leftsided heart diseases, et al.) were found.

Follow-up and Outcome Measures
The primary clinical outcome was the rate of 5-year all-cause mortality. Follow-up information was obtained by telephone calls or routine clinical visits to monitor outcomes every three months during the rst year and every six months after that. Patient deaths were con rmed by medical records or death certi cates review. Overall survival time of patients was measured from the date of CPET until ve years or to the date of patient death, whichever occurred rst.

Statistical Analysis
Continuous and categorical variables were presented as the mean ± standard deviations and counts (proportions), respectively. Comparisons between the survivors and non-survivors were made by twotailed independent samples t-tests for normally distributed variables and nonparametric Mann-Whitney U test for not normally distributed variables, while chi-square test was used for categorical variables. Cox proportional hazards regression analysis was performed to evaluate the prognostic value of different parameters. Variables included in multivariable models were based on their clinical relevance and statistical signi cance in the univariable analysis. To identify the risk thresholds and trends, the continuous association of individual CPET estimates with mortality was modeled using restrictive cubic splines models with four knots at the 5 th , 35 th , 65 th , and 95 th percentiles adjusting for gender (female), WHO-FC, PVR, and N-terminal prohormone of brain natriuretic peptide (NT-proBNP). The Kaplan-Meier survival curves were employed to depict the differences in the 5-year survival rate between patients with values above or below the risk thresholds estimated from cubic splines models. Curves were compared using the log-rank test. ROC analysis was used to evaluate the additional role of CPET parameters in risk prediction compared to European guidelines and identify the optimal prediction models. Correlations between CPET variables and pulmonary hemodynamics were also explored using the Pearson correlation coe cient. A two-sided P<0.05 was considered statistically signi cant. Statistical analyses were performed using R statistical version 3.6.3 (R Project for Statistical Computing) within RStudio statistical software version 1.1.453.

Baseline Characteristics
Of 211 consecutive patients with IPAH who underwent CPET, one was lost to follow-up due to lung transplantation, and the remaining 210 patients (51 men and 159 women, aged 31.9±10.0 years) were included in the nal analysis. The majority presented with WHO-FC II and III, consisting of 105 and 91 individuals. The mean levels of NT-proBNP were 1302.2 pg/ml, with 6MWD 388.0±93.3 meters and the cardiac index 2.8±0.9 ml/min/m 2 . A total of 180 (85.7%) patients received speci c targeted therapy. Patients not receiving targeted therapy were mainly due to nancial burden, intolerability, and cautions about adverse effects. During a median 34-month follow-up, 37 patients died in the entire cohort, accounting for a 17.6 % mortality rate.

Comparisons between Survivors and Non-Survivors
Demographics did not differ between survivors and non-survivors, whereas non-survivors had signi cantly worse WHO-FC (III/IV, 64.9% vs. 42.8%, P=0.007), and decreased 6MWD (352.5±110.4 vs. 396.3±87.4, P=0.042, Table 1). Non-survivors also had higher mPAP, PVR, RAP, and lower cardiac index (all P<0.05). No group differences were observed in CPET-derived parameters for peak work rate, peak respiratory exchange rate, and peak heart rate. Distinctively, non-survivors had lower levels of heart rate reserve, peak VO 2 , peak VO 2 /HR, AT, peak systolic blood pressure (SBP), peak diastolic blood pressure, and had poorer ventilation e ciency (e.g., higher VE/VCO 2 Slope and lower PETCO 2 @AT).

Predictors of 5-Year Mortality
In the univariable Cox regression analysis, WHO-FC, NT-proBNP, hemodynamic parameters, peak work rate, HRR, peak VO 2 , peak VO 2 /HR, peak SBP, AT, PETCO2@AT, VE/VCO 2 slope, and OUES were associated with an increased risk of 5-year mortality (all P<0.05, Table 2). To reduce collinearity between CPET parameters, separate multivariable Cox regression models were developed to assess their independent prognostic signi cance for mortality risk. As there were not enough primary outcomes, thereby avoiding over tting for mortality, we only put gender (female), WHO-FC, PVR, and NT-proBNP into the multivariable analysis considering their clinical signi cance and chi-square values and P values in the univariable analysis. Each line in Table 3 represents a separate model adjusting for gender (female), WHO-FC, PVR, and NT-proBNP. Among these CPET variables, OUES, peak VO 2 /HR, and peak VO 2 remained independently associated with mortality, in descending order of prediction power ( 2 = 37.39 > 35.96 > 35.57). The lower these variables were, the higher the mortality risk (OUES: hazard ratio [HR] (95% con dence interval [CI]) 0.998 (0.996-0.999), P=0.005; peak VO 2 /HR: 0.717 (95% CI 0.534-0.964), P=0.027; and peak VO 2 : 0.868 (95% CI 0.758-0.995), P=0.042). The distribution of their levels in different WHO-FC groups is demonstrated in Fig. 1. As cardiac function worsened, exercise capacity indicated by these parameters gradually declined (P-value for trend <0.05).

Risk Thresholds Exploration
As shown in Fig. 2, optimal exercise tolerance re ected by higher OUES, peak VO 2 /HR, and peak VO 2 were bene cial in reducing mortality risk, yet associations of these metrics with mortality appeared to differ at low to moderate levels. For peak VO 2 , there was a positive linear association, with the lowest mortality risk seen in those with peak VO 2 reaching approximately 20 ml·kg -1 ·min -1 . The adjusted HR of 5- year mortality appeared to continuously escalate as peak VO 2 decreased below levels of 12.2 ml·kg -1 ·min -1 . In contrast, the relationship between OUES and mortality was curvilinear, with maximal mortality between 0.5 and 0.91 among which the risk reaches the peak at 0.72, and with declining mortality risk above 0.91. Peak VO 2 /HR between 1 and 5.3 ml·min -1 ·beat -1 is associated with the highest excess mortality, whereas peak VO 2 /HR ≥5.3 ml·min -1 ·beat -1 corresponds to the threshold above which mortality risk starts to decrease and reach a steady-state above the estimated level of 6 ml·min -1 ·beat -1 (P=0.828 for non-linearity for OUES; P=0.0324 for non-linearity for peak VO 2 /HR).

Long-Term Survival Analysis
Kaplan-Meier curves for survival strati ed by OUES, peak VO 2 /HR, and peak VO 2 are shown in Fig. 3.
Prediction Capacity for Mortality ROC curves are plotted to determine the overall accuracy for mortality risk prediction (Fig. 4). The combination of OUES, peak VO 2 /HR, and peak VO 2  was preferable to the remaining models given its largest AUC and most excellent accuracy for mortality risk prediction (Model 6: AUC=0.838, 95% CI 0.736-0.941; Speci city 89.9%; Sensitivity 72.7%).

Correlations between CPET and Pulmonary Hemodynamics
As the heat map showed (Fig. 5), blue represents positive correlation, whereas red represents negative correlation. The darker the color, the stronger is the correlation. The

Discussion
In this long-term follow-up study of 210 patients with IPAH, the OUES, peak VO 2 /HR, and peak VO 2 were found to be independent markers for 5-year mortality. Notably, we offered a fresh perspective on the risk thresholds and prognostic impact of varying intensities of exercise capacity on mortality. The joint combination of these parameters can further improve the risk prediction capacity compared to contemporary assessment tools for patients with IPAH.
Identifying patients at higher mortality risk of IPAH remains an area of intense investigation and is of utmost importance to guide their therapeutic strategies. A growing body of evidence has shown the prognostic value of exercise tolerance in predicting survival in patients with PAH. For example, peak VO 2 , PETCO 2 @AT, VE/VCO 2 slope, oxygen pulse, and peak SBP during CPET were correlated with survival in patients with PAH [14]. Nonetheless, only peak VO 2 (≤10.4 mL·kg −1 ·min −1 ) and peak SBP (≤120 mmHg) were independent predictors of mortality in the previous studies [6,14]. These ndings were in part in agreement with the results of our study conducted in patients with IPAH. Remarkably, we further enhanced the understanding of the risk threshold (12 ml·kg -1 ·min -1 ) for capturing the dynamic trend in mortality risk. Consistently, exercise capacity is considered a vital component of the REVEAL 2.0 [15] and ESC/ERS risk strati cation tools [16], and current guidelines recommend assessing exercise tolerance for decision-making [1,17]. Accumulating evidence suggests that exercise training confers a protective effect on exercise capacity as indicated by improvement of peak VO 2 [18][19][20], possibly due to the increased capillary density of skeletal muscle [19] and optimized oxidative enzyme function in the peripheral muscles [21]. However, peak VO 2 was not proved as an independent predictor in some other PAH cohort studies [6,22], probably ascribed to a lack of patient motivation and premature termination of exercise by the examiner. Overall, these discrepant ndings could be attributed to the differences in sample size, associated conditions, or comorbidities. In this matter, our ndings may be better applicable to IPAH with fewer comorbidities.
OUES, a useful submaximal exercise index of aerobic tness, represents the absolute rate of increase in oxygen consumption per 10-fold increase in ventilation. A higher OUES represents more oxygen being delivered in the body, while a lower OUES represents an enormous amount of ventilation required for given oxygen uptake. Three main factors can in uence the values of OUES, including arterial carbon dioxide setpoint, carbon dioxide production, and death space ventilation [23,24]. An earlier investigation showed that the arterial carbon dioxide setpoint did not vary between normal individuals and HF during exercise [25]. As such, both systemic perfusion (related to carbon dioxide production) and pulmonary perfusion (related to dead space ventilation) are related to the magnitude of OUES. Previous literature has shown that OUES was strongly correlated with peak VO 2 and provided better predictive power over peak VO 2 in patients with left-sided HF [11,26] and PAH [10]. These observations were consistent with our ndings regarding the superiority of OUES over peak VO 2 in predicting adverse outcomes.
Peak VO 2 /HR re ects the amount of oxygen consumed per heartbeat (i.e., the product of stroke volume and arteriovenous oxygen difference during exercise). It indirectly indicates the maximal myocardial oxygen supply and cardiac function reserve under stress. The correlation analysis supports that peak VO 2 /HR is associated with decreased cardiac index and a higher mortality risk beyond the risk threshold. In comparison, a recent study explored the role of peak VO 2 /HR in patients with chronic obstructive pulmonary disease and concluded that it might be an indicative parameter of lung hyperin ation, PH, and HF comorbidity [27]. Apart from this, we further supported the role of peak VO 2 /HR in estimating long-term mortality risk in patients with IPAH. It emphasizes maintaining an appropriate cardiovascular function by individualized exercise training and aggressive speci c drug treatments.
Meanwhile, interestingly, we found that VE/VCO 2 slope did not predict the poor outcome, in contrast to the previous literature [9,12,14,28], but consistent with others [6,8,10,11]. For example, Davies et al. [11] concluded that OUES rather than peak VO 2 and VE/VCO 2 slope was the most powerful predictor in patients with left-sided HF. Similar ndings were also observed in patients with PAH [8,10]. We proved that OUES provided better prognostic information than that provided by VE/VCO 2 , and more meaningfully, shed light on their risk thresholds. This may be linked to the fact that OUES re ects not only the status of pulmonary perfusion but systemic perfusion. Nonetheless, opposite results showing that VE/VCO 2 plays a predominant role in predicting outcomes were observed in a study by Arena et al. [12]. Speculatively, underlying clues explaining these discrepancies include the -blockers use, the study period, and the population inhomogeneity. Moreover, the 6MWD in our study did not associate with the outcome, which may be attributable to the selection of stable patients with relatively high values where peak VO 2 , peak VO 2 /HR, and OUES become more sensitive to functional exercise capacity.
The OUES seemed to be a vital predictive factor superior to the other CPET-derived parameters. There were curvilinear relationships between OUES and peak VO 2 /HR and mortality risk. The joint combination of independent indicators -OUES, peak VO 2 /HR, and peak VO 2 had a better prognostic prediction value than the simpli ed risk estimates in 5-year mortality. Close monitoring of these CPET variables may aid in better risk strati cation and individualized therapy for patients with IPAH. The OUES and peak VO2/HR seem likely to further increase mortality risk at some low measurement levels. However, However, future studies are needed to explore the underlying exacerbating mechanisms and elaborate whether cardiopulmonary rehabilitation could serve as a critical strategy to alleviate exercise intolerance, ameliorate disease progression, and improve outcomes in patients with IPAH.
This long-term follow-up study characterized the continuous relationship between exercise tolerance and mortality risk. Greater statistical power using adjusted restrictive cubic splines models enhanced detection of the different associations seen with CPET risk thresholds and mortality, which have been rarely undertaken before. Several limitations also warrant discussion. First, our study only included patients with IPAH with a lower percentage of WHO-FC IV. The ndings thereby may not be generalized to other PAH and those with poorer cardiac function. Second, although the data were prospectively collected, our analysis was retrospective in nature. Despite the multivariate adjustment, we cannot exclude the possibility of residual confounding and reverse causality. Finally, limited information was available on the PAH-speci c treatments on the changes of the CPET variables. Future multi-center and large-scale studies are required to validate our ndings and explore the potential aggressive therapeutic effects on exercise tolerance and disease outcomes.

Conclusions
Suboptimal exercise tolerance re ected by OUES, peak VO 2 /HR, and peak VO 2 under certain thresholds provides excellent prognostic information for 5-year mortality in patients with IPAH. The joint combination of these parameters and simpli ed risk estimates can improve the risk prediction capacity compared to contemporary assessment tools for patients with IPAH.

Consent for publication
Not applicable

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests. Authors' contributions ZH and LY contributed to the study design, data analysis, manuscript drafting and revision. ZZ, QZ acquired data, critically reviewed, and revised the manuscript. YT, QJ, YZ, XL, AD, MH performed literature search. QL and ZL provided professional advice on data interpretation, critically reviewed, and revised the manuscript. All authors contributed substantially to the work and approved the nal manuscript.    PVR, and NT-proBNP. AT, anaerobic threshold; CI, con dence interval; HR, heart rate; HRR, heart rate reserve; NT-proBNP, N-terminal prohormone of brain natriuretic peptide; OUES, oxygen uptake e ciency slope; PETCO 2 @AT, end-tidal partial pressure of carbon dioxide at anaerobic threshold; PVR, pulmonary vascular resistance; SBP, systolic blood pressure; VCO 2 , carbon dioxide output; VE, minute ventilation; VO 2 , oxygen consumption; WHO-FC, World Health Organization Functional Class; WR, work rate. Figure 1 Boxplots of OUES (A), peak VO 2 /HR (B), and peak VO 2 (C) in WHO-FC groups. OUES, oxygen uptake e ciency slope; VO 2 , oxygen consumption; HR, heart rate; WHO-FC, World Health Organization Functional

Figure 2
Restrictive cubic splines of associations between 5-year mortality risk and OUES (A), Peak VO 2 /HR (B), and Peak VO 2 (C). The restrictive cubic splines models were adjusted for gender, WHO-FC, PVR, and NT-proBNP, with knots at the 5 th , 35 th , 65 th , and 95 th percentiles. The horizontal dash line indicates a null hazard ratio equal to 1. The solid line denotes the estimated hazard ratio. The shaded area represents the 95% con dence limits. The hazard ratio is per each absolute increase of 1 unit of individual CPET parameter level. HR, heart rate; NT-proBNP, N-terminal prohormone of brain natriuretic peptide; OUES, oxygen uptake e ciency slope; PVR, pulmonary vascular resistance; VO 2 , oxygen consumption; WHO-FC, World Health Organization Functional Class.