Cardiopulmonary Exercise Testing Characterizes Silent Cardiovascular Abnormalities in Asymptomatic Pediatric Cancer Survivors

Late-onset cardiovascular complications are serious concerns for pediatric cancer survivors (PCS) including those who are asymptomatic. We investigated whether cardiopulmonary exercise testing (CPET) can delineate the underlying pathophysiology of preclinical cardiovascular abnormalities in PCS. We examined CPET data via cycle ergometer in asymptomatic PCS with normal echocardiogram and age-matched controls. Peak and submaximal parameters were analyzed. Fifty-three PCS and 60 controls were studied. Peak oxygen consumption (VO2), peak work rate (WR), and ventilatory anaerobic threshold (VAT) were significantly lower in PCS than controls (1.86 ± 0.53 vs. 2.23 ± 0.61 L/min, 125 ± 45 vs. 154 ± 46 W, and 1.20 ± 0.35 vs. 1.42 ± 0.43 L/min, respectively; all p < 0.01), whereas peak heart rate (HR) and ventilatory efficiency (a slope of minute ventilation over CO2 production or ∆VE/∆VCO2) were comparable. Peak respiratory exchange ratio (RER) was significantly higher in PCS (p = 0.0006). Stroke volume (SV) reserve was decreased in PCS, indicated by simultaneous higher dependency on HR (higher ∆HR/∆WR) and lower peak oxygen pulse (OP). Twelve PCS with high peak RER (≥ 1.3) revealed lower pVO2 and VAT than the rest of PCS despite higher ventilatory efficiency (lower ∆VE/∆VCO2), suggesting fundamental deficiency in oxygen utilization in some PCS. Poor exercise performance in PCS may be mainly attributed to limited stroke volume reserve, but the underlying pathophysiology is multifactorial. Combined assessment of peak and submaximal CPET parameters provided critical information in delineating underlying exercise physiology of PCS.


Introduction
Recent remarkable progress in cancer treatment has enabled nearly 80% of pediatric cancer patients to survive into adulthood cancer free [1]. At the same time, long-term cardiovascular complications have become concerns for pediatric cancer survivors (PCS) [2,3]. The pathobiology of progression of cardiovascular complications in PCS has been extensively studied, and multiple pathological mechanisms have been proposed. However, the clinical manifestations of long-term cardiovascular complications remain poorly understood [4]. Routine echocardiographic surveillance is recommended by multiple clinical guidelines to screen atrisk patients [5], but its reliability in detecting subtle preclinical cardiovascular impairment is limited.
Cardiopulmonary exercise testing (CPET) is a useful, noninvasive method to assess cardiopulmonary fitness level in children and adolescents with heart disease. Unlike echocardiography and magnetic resonance imaging (MRI), which primarily assess cardiac function at the resting condition, CPET mainly measures the functional reserve of all organs involved in exercise performance, including heart, lung, blood, vasculature, and skeletal muscles, in which cardiac output accounts for approximately 70% of total exercise 1 3 performance. Several studies have shown that PCS have a reduced exercise performance compared with age-matched normal controls [6][7][8][9], but the underlying pathophysiology of reduced exercise performance remains elusive.
In this study, we investigated overall exercise performance by combinational assessment of submaximal CPET parameters that represent a dynamic trend during intermediate phase of exercise and conventional peak CPET parameters. We hypothesized that the underlying mechanisms of poor exercise performance in PCS can be delineated by this new combinational approach with submaximal and peak CPET parameters.

Patients and Methods
A retrospective chart review of CPET data from the database of the Exercise Laboratory, Nemours Cardiac Center, Nemours Children's Health, Wilmington, DE, was conducted from 2017 to 2020. The study was approved by the Institutional Review Board of the hospital.

Patients
We retrospectively studied asymptomatic PCS followed at the Cancer Survivorship Program at Nemours Children's Health, Delaware, who were referred for CPET to assess physical fitness levels. The inclusion criteria consisted of the following: (1) age ≥ 10 years, (2) off-cancer treatment ≥ 1 year, (3) intact musculoskeletal system and neurological function, (4) body mass index (BMI) < 30 kg/m 2 , and (5) left ventricular shortening fraction (LVSF) ≥ 28% or left ventricular ejection fraction (LVEF) ≥ 55% by echocardiography. Age, sex, height, weight, and BMI of the patients were collected at the time of CPET. For PCS, primary diagnosis, age at diagnosis, cumulated dosage of anthracycline (mg/ m 2 ), and history of radiation therapy were recorded. Sex-, age-, and weight-matched control patients were recruited from the database of the Exercise Laboratory at the Nemours Cardiac Center.

Cardiopulmonary Exercise Testing (CPET)
The study was performed on bike ergometer (VIA Sprint 150 P, Yorba Linda, CA) following RAMP protocol with 10-25-W (W)/min increments up to peak exercise (approximately 0.3 W/kg/min). Heart rate and oxygen saturation (SaO 2 ) were continuously monitored by standard 12-lead electrocardiogram (ECG) and pulse oximeter, respectively. Blood pressure was measured every 2-3 min during exercise and recovery phases. Oxygen consumption (VO 2 ), carbon dioxide production (VCO 2 ), and minute ventilation (VE) were measured continuously during all exercise testing using a calibrated metabolic measurement system (Vmax Sensor Medics, Palm Springs, CA). The exercise protocol was continued until the patient stopped because of symptomatic limitations. Achievement of peak exercise level was confirmed by either peak HR of more than 90% of estimated maximum HR for age (220 -age) or respiratory exchange ratio (RER) of 1.1 or higher.
Peak and submaximal exercise parameters were obtained in combination with continuous monitoring of vital signs and ECG recording. Peak values of HR (pHR), VO 2 (pVO 2 ), VCO 2 (pVCO 2 ), oxygen pulse (pOP), work rate (pWR), minute ventilation (pVE), and peak RER (pRER) were measured. Submaximal CPET parameters consist of ventilator anaerobic threshold (VAT) and submaximal slope parameters, including ∆VO 2 /∆WR (an inverse of work efficiency), ∆HR/∆WR (heart rate dependency), ∆VO 2 /∆HR (a surrogate of stroke volume), oxygen uptake efficiency slope (OUES; a slope in a relationship between logarithm of VE and VO 2 ) [10], and ∆VE/∆VCO 2 (an inverse of ventilatory efficiency). Submaximal CPET slopes represent a physiological trend of how exercise parameters change in response to the intermediate phase of programmed incremental exercise up to the anaerobic threshold (AT) (Suppl. Fig. 1). All CPET parameters were presented as an absolute value. Peak VO 2 , VAT, WR, ∆VO 2 /∆HR, and ∆HR/∆WR were also presented as relative values indexed by body weight (*indicates that the values were indexed by body weight).

Statistics
Distribution of patients' demographics as well as peak and submaximal parameters were compared between PCS and control groups. Mean and standard deviation (SD) for continuous variables and count and percentage for categorical variables are reported. Two-sample t test and χ 2 test were used to compare the mean and proportion, respectively, between two groups. Analysis of covariance (ANCOVA) was used to compare the regression lines between two groups. Model/test assumptions were checked before data analysis. All tests were two-tailed at the level of significance of 0.05. The statistical software R (version 3.5.2: R Core Team) was used for data analysis.

Clinical Profile of PCS
Fifty-three PCS (26 male and 27 female) with various malignancies were studied ( Table 1). The ages at diagnosis of malignant diseases and CPET were 6.4 ± 4.5 years and 14.5 ± 2.8 years, respectively. All patients received anthracycline for chemotherapy. Total cumulated anthracycline dosage was 235 ± 102 [60-450] mg/m 2 . Sixteen patients received radiation therapy: 7 chest/mediastinum, 3 flank/lumbar region, 2 total body irradiation, and 4 others. Echocardiogram revealed LVSF 33.9 ± 3.1%, LVEF 61.7 ± 5.8%, and left ventricular myocardial index (LVMI) 28.9 ± 6.9 (normal < 39). Sixty age-, sex-, and weightmatched control patients were selected. Table 2 summarizes the demographic profile and results of CPET. Although weight and BMI were comparable between PCS and control groups, the height of PCS was significantly lower than controls. Peak HR was comparable between the two groups. Peak exercise parameters, pVO 2 , pOP, and pWR, were significantly lower in PCS than in controls. Peak RER was significantly higher in PCS than in controls, but pVE was comparable. Pediatric cancer survivors showed significantly lower VAT, OUES, and ∆VO 2 /∆HR than controls. However, the difference was not statistically significant when VAT and ∆VO 2 /∆HR were indexed by weight. Heart rate dependency (∆HR/∆WR) was significantly higher in PCS than in controls in both absolute-and weight-indexed values. Work efficiency (∆VO 2 /∆WR) and ventilatory efficiency (∆VE/∆VCO 2 ) were comparable between the two.

Body Mass and CPET Parameters
Relationships between the weight and exercise parameters in absolute CPET values were examined (Fig. 1). In controls, a strong positive relationship was demonstrated between weight and CPET parameters, including pVO 2 (r = 0.73), pWR (r = 0.74), and OUES (r = 0.59), suggesting exercise performance is proportional to body weight. In PCS, not only were the positive relationships weaker (r = 0.39-0.58), but each slope was more gradual than in controls, indicating that the PCS group consisted of more diverse body composition and that body weight in PCS did not contribute to the increase in CPET parameters as effectively as in controls. However, these differences did not reach statistical significance.

Chronotropic Function and Stroke Volume Reserve
As shown in Table 1, ∆HR/∆WR was significantly higher and pOP and pVO 2 were significantly lower in PCS than controls. With simultaneous presentation of ∆HR/∆WR  and pOP as well as ∆HR/∆WR and pVO 2 , a combination of lower pOP or pVO 2 and higher ∆HR/∆WR was more prominent in PCS, suggesting that the increased HR dependency is more noticeable in PCS probably due to limited stroke volume reserve. The slopes of a correlation line of ∆HR/∆WR-pOP and ∆HR/∆WR-pVO 2 were statistically significant between PCS and controls (Fig. 2).

Submaximal Exercise Parameters to Predict Peak Exercise Performance
An excellent correlation was demonstrated between submaximal parameters (VAT, OUES, and ∆VO 2 /∆HR) and pVO 2 with no significant difference between PCS and control groups, suggesting these submaximal parameters can reliably predict peak pVO 2 equally in both groups (Fig. 3). However, the distribution is markedly different between PCS and controls on the almost identical correlation lines; PCS revealed lower values in both x-and y-axes, as also shown in Table 2. The data suggest that there is no difference in exercise persistence in a partially anaerobic condition up to peak exercise between PCS and controls but that overall exercise performance was significantly lower in PCS.

Ventilation and Peak Exercise Performance
Ventilatory efficiency in O 2 uptake and CO 2 elimination at the peak exercise was examined by the correlations of pVE-pVO 2 and pVE-pVCO 2 , respectively, which demonstrated excellent positive linear correlations in both PCS and controls (Fig. 4). There was no statistically significant difference in the correlation lines between PCS and controls (p = 0.62 and 0.52 for pVE-pVO 2 and pVE-pVCO 2 , respectively), Fig. 1 Body mass effects. Correlations between weight (kg) and exercise parameters, including peak oxygen consumption (pVO 2 ) (L/min), peak work rate (pWR) (Watt), and oxygen uptake efficiency slope

Higher Peak RER in PCS
Peak RER (= pVCO 2 /pVO 2 ) was significantly higher in PCS than in controls. Although both pVCO 2 and pVO 2 were significantly lower in PCS than in controls, higher pRER in PCS may indicate disproportionally low pVO 2 as an essential feature of PCS rather than excessive exercise effort or hyperventilation. When VAT* was plotted with pRER, there was a negative correlation in both groups (Fig. 5A). Compared with controls, PCS had an extreme group with higher RER and lower VAT*. Lower VAT* indicates earlier initiation of anaerobic metabolism. When PCS were divided into higher pRER (≥ 1.3; n = 12) and lower pRER (< 1.3; n = 41) subgroups (Fig. 5B), the higher pRER group showed significantly lower pVO 2 *, VAT, and VAT* with higher ventilatory efficiency (lower ∆VE/∆VCO 2 ) than the lower pRER group (Fig. 5C), suggesting that the higher RER group in PCS is characterized by lower aerobic capacity and lower overall exercise performance with enhanced ventilator efficiency.

∆VO 2 /∆WR and Possible Cardiovascular Risk in PCS
A correlation between pWR and pVO 2 was excellent in both PCS and controls, and the two correlation lines were almost identical (Fig. 6A). Although ∆VO 2 /∆WR was comparable between PCS and controls (Table 1), PCS consisted of two outlier groups, including (a) higher ∆VO 2 /∆WR and relatively lower pVO 2 * (n = 4) and (b) markedly lower ∆VO 2 /∆WR and significantly low pVO 2 * (n = 4) compared with the rest (n = 45), when ∆VO 2 /∆WR and pVO 2 * are plotted simultaneously as shown in Fig. 6B. The remaining group of PCS nearly overlaps with controls. The presence of these two extreme groups indicates a diverse clinical phenotype of PCS. Group a was characterized by comparable or slightly lower exercise performance than the rest of PCS but with significantly higher ∆HR/∆WR* (HR dependency) and higher ∆VO 2 /∆WR (= low work efficiency) than the rest. On the other hand, group b showed significantly lower pVO 2 *  and lower ∆VO 2 /∆WR than the rest, suggesting intrinsically limited VO 2 increase in response to exercise rather than high work efficiency. These data suggest substantial heterogeneity regarding underlying pathology in exercise performance in PCS as a preclinical cardiovascular abnormality (Fig. 6C).

Discussion
Our current study demonstrated that asymptomatic PCS presented with significantly diminished exercise performance than controls despite normal global LV systolic function at rest and that the causes of their poor exercise performance are likely multifactorial. The simultaneous presentation of ∆HR/∆WR and pOP indicated that PCS in general had higher dependency on HR increase than increase of OP at the peak exercise, suggesting primary limitation in stroke volume reserve. Peak RER (= pVCO2/pVO2) was significantly higher in PCS than in controls, and a group of PCS with higher pRER (≧ 1.3) revealed significantly lower pVO2* than PCS with lower pRER (< 1.3), suggesting limited aerobic capacity in some PCS. There were two small outlier groups of abnormal exercise performance in PCS with different underlying mechanisms. These findings represent the significant heterogeneity of abnormal cardiovascular presentation in PCS, which was identified by our novel CPET analysis.

Decreased Exercise Performance in Asymptomatic PCS
Anthracycline-induced cardiotoxicity is a major cause of late cardiovascular complications in PCS that occur decades after the initial treatment [11]. Late-onset cardiotoxicity is insidious and nonspecific yet progressive and irreversible [12]. Thus, early recognition of cardiotoxicity is essential to protect patients from developing symptomatic cardiomyopathy or advanced heart failure. Reliability of echocardiography in predicting late cardiovascular complications is limited as normal myocardial status in younger ages may not be completely free from late cardiotoxicity [13,14]. Long-term cardiovascular complications for PCS not only pertain to direct myocardial dysfunction and heart failure but also include increased incidence of coronary artery disease, stroke, and variable vascular diseases [15]. Early recognition and management of preclinical cardiac dysfunction are critical in optimizing survival and improving quality of life of PCS [3,16,17].
In this study, we demonstrated that peak exercise performance values including pVO 2 , pOP, and pWR were significantly lower in PCS than controls, in agreement with the previous published studies [6][7][8][9][18][19][20][21]. Two recent studies demonstrated no significant difference in parameters obtained by stress echocardiogram between PCS and agematched controls at peak exercise [22,23], suggesting that markers of ventricular myocardial performance may not always be a sensitive marker of preclinical cardiovascular abnormality and that other peripheral factors, skeletal muscle alteration and/or vascular dysfunction, may also contribute to a reduced exercise performance. Ness et al. studied 1041 adult survivors of childhood cancer and demonstrated high incidence of exercise intolerance (63.8%) due to a combination of cardiac, pulmonary, autonomic nervous system-mediated, and peripheral muscular impairment [24]. A similar trend has been presented in adult breast cancer survivors with preserved LVEF, in which impaired peripheral vascular function and skeletal muscle dysfunction were attributed to decreased pVO 2 in addition to impaired cardiac function [25]. Worsening of exercise efficiency (high ∆VO 2 /∆WR) may represent vascular dysfunction commonly seen in elderly people because of loss of vascular elasticity (vascular senescence) [26]. Premature aging either by DNA damage or telomerase shortening in the cardiovascular system is suggested as a cause of increased incidence of cardiovascular events in PCS [27].

Possible Mechanisms of Preclinical Cardiotoxicity Characterized by CPET
From our current study, we propose certain underlying mechanisms responsible for the poor exercise performance in PCS. First, reduced pOP in PCS was noted in combination with preferential increase in HR in PCS (Fig. 2). A limited stroke volume reserve with higher dependency on HR increase was previously reported in a small group of asymptomatic PCS [28]. With an exercise MRI study, Foulkes et al. demonstrated that reduced peak exercise performance in PCS was associated with decreased stroke volume reserve and cardiac index [29]. A combination of low pVO 2 and high pRER was noted to have significantly higher mortality in adult patients with chronic heart failure [30], suggesting that an impaired VO 2 increase is a fundamental abnormality in some PCS. In contrast, relatively lower pRER was noted during intense exercise in well trained athletes than in sedentary controls [31]. Collectively, a reduced peak oxygen delivery/consumption is likely to be a central pathophysiology in PCS regardless of identifiable global ventricular dysfunction, underscoring the critical importance of CPET in risk-stratifying asymptomatic PCS. There was no significant difference in exercise persistence between PCS and controls based upon comparable trends between submaximal parameters (VAT, OUES, and ∆VO 2 /∆HR) and peak VO 2 (Fig. 3).
Second, there may be a difference in the composition of metabolically active skeletal muscle mass between PCS and controls, as shown in Fig. 1, although we were not able to demonstrate statistical significance. Sarcopenia and skeletal muscle dysfunction are known complications after cancer treatment [32,33]. A difference in capillary density and mitochondria concentration within the myocytes can also affect oxygen utilization at a tissue level [34]. Repetitive skeletal muscle contraction is also known to augment venous return and thus cardiac output (muscle pump). It is plausible that PCS are more prone to inactive lifestyles responsible for physical deconditioning [35]. Altered peripheral oxygen utilization may contribute to poor exercise performance despite normal ventricular systolic function in some PCS.
Lastly, there were small subgroups of outliers of high ∆VO 2 /∆WR and low ∆VO 2 /∆WR (group a and b in Fig. 6B, respectively). High ∆VO 2 /∆WR implies a high oxygen cost to perform external work, commonly seen in obesity, use of additional muscles, or recruitment of less efficient muscle fibers [36]. These people showed comparable exercise performance with the main PCS but had decreased peak WR* and significantly increased HR dependency similar to group b. A combination of low ∆VO 2 /∆WR and low pVO 2 indicates decreased cardiac output frequently seen in patients with chronic heart failure [37] and should be regarded as a high-risk group. Although obese subjects were not included in this study, it is plausible that some PCS with decreased skeletal muscle mass and presumably increased body fat present with elevated ∆VO 2 /∆WR.

Significance of Submaximal CPET Parameters in Assessing Exercise Performance
Peak CPET parameters have been considered as a gold standard in assessing cardiac reserve. However, this prediction may be hampered when patients cannot reach maximal exercise status due to multiple reasons, including physical deconditioning, emotional discomfort, or lack of compliance [38]. Submaximal slope parameters represent dynamic physiological responses under aerobic metabolism and are independent of patients' effort. Even if the maximal exercise is not fulfilled, submaximal CPET data could provide ample information regarding underlying cardiac and metabolic profiles that support exercise [39]. Certain submaximal parameters, VAT, ∆VO 2 /∆HR, and OUES, are known to correlate well with pVO 2 and are easily obtained from the routine CPET worksheet, even retrospectively. Despite a wealth of information, they have been underutilized or only used sporadically in conventional pediatric CPET analysis. Systematic incorporation of submaximal CPET parameters into routine peak parameters would augment our understanding of exercise physiology in more depth [40][41][42].

Application of Simultaneous Assessment by Two CPET Parameters
We have characterized a new method of assessing CPET data by combining peak and submaximal parameters (including weight) simultaneously to compare the trends of two groups ("Two-dimensional CPET Analysis"). A simultaneous assessment of ∆HR/∆WR and pOP (= a surrogate of stroke volume at peak exercise) or pVO 2 showed not only a good inverse relationship between the two parameters but also demonstrated a clear difference between PCS and controls (Fig. 2). Submaximal parameters, VAT, OUES, and ∆VO 2 /∆HR, were plotted with pVO 2 in both groups, which showed an excellent positive correlation with almost identical correlation lines, suggesting that these submaximal parameters are reliable markers to predict peak exercise performance in both groups (Fig. 3). Two-dimensional analyses by ∆VO 2 /∆WR and pVO 2 * further identified two distinctive outliers in PCS: one with probably inefficient peripheral energy production (group a) and the other with limited ventricular myocardial reserve with lower aerobic capacity (group b) (Fig. 6). The two-dimensional CPET correlation analysis is easy to perform even retrospectively from any existing standard exercise worksheet and provides substantial additional information to interpret baseline exercise physiology without extra investment [42].

Limitations
There are several limitations in our study that need to be addressed. This is a retrospective study with a relatively small sample size in a single center. The PCS group represents a heterogeneous population regarding primary diagnosis, cumulated dosage of anthracycline, years at diagnosis, years after remission, body habitus, and the level of baseline physical activities. Notably, physical conditioning was not specifically addressed in either group. Skeletal muscle mass was not directly measured, which could affect the interpretation of CPET results. There may be a selection bias as PCS included in this study were those who were willing to undergo CPET for functional assessment of their exercise performance. We also excluded obese PCS from the study primarily to optimize the CPET interpretation as obesity may be an important pathological feature in PCS. Despite these limitations, our current study clearly underscores the primary involvement of the reduced stroke volume with heterogeneous abnormalities other than direct myocardial impairment in otherwise asymptomatic PCS.

Conclusion
Our novel approach in combining peak and submaximal CPET parameters enabled us to delineate multiple contributing factors involved in decreased exercise performance in PCS, which makes CPET an important surveillance tool in risk-stratifying future cardiovascular complications in PCS. This simple unique CPET application is useful and physiologically relevant and should be used more routinely in following PCS. Regular exercise is proven to be beneficial in attenuating progression of anthracycline-induced cardiotoxicity in both human and animal studies [43,44]. The decrease in exercise performance without myocardial dysfunction may still imply increased future cardiovascular risk. Whether these CPET abnormalities during pediatric ages predict serious cardiovascular complications in adulthood decades after the completion of cancer treatment remains to be investigated.