Reduced exercise capacity is associated with left ventricular systolic dysfunction in long‐term survivors of allogeneic hematopoietic stem‐cell transplantation

Abstract Purpose Exercise intolerance is a common complication in survivors of allogeneic hematopoietic stem‐cell transplantation (allo‐HSCT). The aim of this study was to determine if cardiac function measured with echocardiography is associated with exercise capacity measured with cardio‐pulmonary exercise tests in long‐term survivors treated in their youth with allo‐HSCT. Methods The study included 96 patients, of which 54.2% were female, aged 34.9 ± 11.6 years and 17.7 ± 9.3 years after allo‐HSCT. Reduced exercise capacity was defined as <85% of predicted‐peak oxygen uptake (VO2peak). Linear regression was used in the prediction of VO2peak (ml/kg/min). Receiver operating characteristic evaluated the accuracy of predicting reduced exercise capacity. Results VO2peak was 36.2 ± 7.7 ml/kg/min and 43 (44.8%) had reduced exercise capacity. Left ventricular ejection fraction was 55.4 ± 5.9% and global longitudinal strain (GLS) was −17.6% ± 2.0%. Left and right ventricular functions were significantly lower in survivors with reduced exercise capacity. Increased body mass index, lower physical activity score, reduced pulmonary function (by forced expiratory volume in 1‐s) and reduced left ventricular systolic function (by GLS) were significant independent predictors for reduced VO2peak. GLS was superior to other echocardiographical indices for identifying reduced exercise capacity (area under curve = 0.64, p = 0.014). Conclusions Left ventricular systolic dysfunction measured by GLS is associated with reduced exercise capacity in long‐term allo‐HSCT survivors.

Results: VO 2peak was 36.2 ± 7.7 ml/kg/min and 43 (44.8%) had reduced exercise capacity. Left ventricular ejection fraction was 55.4 ± 5.9% and global longitudinal strain (GLS) was À17.6% ± 2.0%. Left and right ventricular functions were significantly lower in survivors with reduced exercise capacity. Increased body mass index, lower physical activity score, reduced pulmonary function (by forced expiratory volume in 1-s) and reduced left ventricular systolic function (by GLS) were significant independent predictors for reduced VO 2peak . GLS was superior to other echocardiographical indices for identifying reduced exercise capacity (area under curve = 0.64, p = 0.014).
Conclusions: Left ventricular systolic dysfunction measured by GLS is associated with reduced exercise capacity in long-term allo-HSCT survivors.

K E Y W O R D S
cardio-pulmonary exercise test, echocardiography, global longitudinal strain, left ventricular ejection fraction, peak ventilatory oxygen-uptake, three-dimensional echocardiography

| INTRODUCTION
Allogeneic hematopoietic stem-cell transplantation (allo-HSCT) is increasingly being selected as a potentially curative therapy for young recipients with malignant and non-malignant disease. 1 In parallel with advances in treatment, more recipients are surviving the initial years after transplantation. 2 Consequently, increased focus has been directed towards improving quality of life by reducing therapy related complications.
A common complication in survivors of cancer therapies is exercise intolerance. The main limiting factors for exercise capacity are cardiac and pulmonary function, hematological capacity and metabolism in skeletal muscle. In long-term survivors of allo-HSCT, the risk of myocardial and pulmonary disease is elevated due to chemotherapy, high rates of cardiovascular risk factors, graft-versus-host disease (GVHD) and physical de-conditioning. A challenge for the clinician is to identify cardiac dysfunction as the cause and to promptly initiate treatment to prevent irreversible heart failure. Cardio-pulmonary exercise tests (CPET) is a widely used method to differentiate reasons for dyspnea. In addition, the acquirement of peak oxygen uptake (VO 2peak ) provides valuable prognostic information on cardiovascular related and all-cause mortality. [3][4][5] Echocardiography provides confirmation of cardiac dysfunction and insight into mechanisms for reduced exercise capacity. However, relationships between VO 2peak and left ventricular (LV) systolic function by echocardiography have been inconsistent. This is in part due to inadequacies with traditional measurements of left ventricular ejection fraction (LVEF). Three-dimensional (3D) imaging and global longitudinal strain (GLS) from speckle tracking echocardiography (STE) have shown to improve sensitivity in detecting subtle effects of cardiotoxicity. [6][7][8][9][10] As such, these methods may align better with observations of oxygen uptake. In particular, is the reported ability of GLS to predict functional capacity in patients with myocardial dysfunction. 11,12 Previous examinations in this cohort have shown a high prevalence of left ventricular systolic dysfunction (LVSD) that is strongly associated with first-line anthracycline therapies. 13 The present study aims to determine if cardiac function measured with echocardiography is associated with exercise capacity measured with CPET, in longterm survivors of allo-HSCT treated in childhood, as adolescents or young adults (CAYA). We hypothesize that modern techniques such as 3D-LVEF or GLS are more accurate in determining relationships with the prognostic marker of VO 2peak (L/min/kg).

| Study design
This nationwide cross-sectional study was designed to include all survivors of allo-HSCT conducted at our institution in a multidisciplinary study investigating the long-term effects of allo-HSCT. Eligibility criteria were: Treatment at our national center for allo-HSCT, age <30 years at transplantation, age >16 years at study inclusion and >5 years follow-up time. Indications for allo-HSCT were malignant and non-malignant diseases. Survivors with Hurler syndrome were excluded due to the possibility of multi-organ pathology as part of their primary disease. Written informed consent was obtained from all participants, and the study was approved by the Regional Committee for Medical and Health Research Ethics.

| Clinical assessment
All participants underwent a medical examinations, questionnaires and blood sampling from June 2014 to February 2016. Dyspnea was classified according to the New York Heart Association (NYHA). 14 Anthracycline cumulative dosage was converted to isotoxic doses of doxorubicin. 15 Blood pressures were acquired after echocardiography (>30 min), in the supine position as the average of three measurements. Blood samples were collected after overnight fasting and analyzed at the hospital laboratory. N-terminal pro-brain-type natriuretic peptide (NT-proBNP) concentrations were determined by an electrochemiluminescence immunoassay (Roche Diagnostics, Basel, Switzerland). The lowest detectable level was 5n/L and manufacturer's recommendations were used for classifying elevated NT-proBNP according to the age and sex specific cutoffs. Anemia was defined as reduced hemoglobin (males: <13.5 g/dL and females: <12 g/dL). Hypertension was defined as use of anti-hypertensive drugs and/or systolic blood pressure (SBP) >140 mmHg or diastolic blood pressure (DBP) >90 mmHg. Hypercholesterolemia was defined as low-density lipoprotein (LDL) >4.1 mmol/L (160 mg/dl) or use of lipid lowering medication. Diabetes mellitus was identified by hemoglobin HBA1c >6.5% (48 mmol/mol), fasting glucose ≥7.0 mmol/L or current use of glucose-lowering medication. Obesity was classified as body mass index (BMI) ≥30 kg/m 2 . Acute graft-versus-host disease (aGVHD) was graded by Glucksberg scales and chronic GVHD (cGVHD) was graded by Shulman scales. 16,17

| Physical activity
Physical activity was quantified from a self-reported questionnaire (HUNT 2/3) that has been validated against measurements of VO 2peak , METS calculations and international physical activity questionnaires in a comparative population. 18 This calculated physical activity during a week as the product of weighted scores for the categories of frequency (scores of 0, 0.5, 1, 2.5, 5.0 ranging from 'never' to 'almost every day'), intensity (scores of 1.0, 2.0, 3.0 ranging from 'light exercise' to 'near-exhaustion') and duration (scores of 0.1, 0.38, 0.75, 1.0 ranging for '<15 min' to '>60 min'). 18 The range is 0 to 15 and higher values reflect greater weekly physical activity. An example: Exercising 2-3 times a week, for a total 60-180 min at a moderate intensity gives a physical activity score of 3.75. This scoring system generates a numerical scale that can be used to quantify the level of physical activity.

| Pulmonary function
Spirometry was conducted as recommended by European Respiratory Society. 19 Pulmonary function in this cohort has previously been described. 20 For the present study, we used forced expiratory volume in 1-s (FEV 1 ) to represent lung function. The rational was two-fold; FEV 1 is readily attainable, and is strongly associated with bronchiolitis obliterans syndrome (BOS), which is the most clinically relevant respiratory disorder in this cohort. Percent of predicted-FEV 1 was calculated for each individual using recommended equations that adjust for race, ethnicity, sex and height. 21 BOS was defined according to the National Institutes of Health (NIH) Consensus Criteria. 22 2.5 | Cardio-pulmonary exercise test CPET was conducted after echocardiography by experienced personnel at our institution. The test was performed on a treadmill (TechnoGym Runrace) using the modified Balke protocol. 23 Incremental changes in ventilatory parameters were measured at regular intervals with Vyntus-CPX (CareFusion). Predicted-VO 2peak and percent of predicted-VO 2peak were calculated for each individual based on equations that adjust VO 2peak by age and sex in a healthy control population. 24 Reduced exercise capacity was defined as VO 2peak < 85% of predicted as recommended. 25 Oxygen-pulse was calculated by dividing VO 2peak by maximal heart rate, and predicted oxygen-pulse by dividing percent of predicted-VO 2peak by maximal heart rate. 25

| Statistical analysis
Statistical analysis was conducted with SPSS version-25 (SPSS, Inc.) and p < 0.05 was considered significant. Histograms and Shapiro-Wilk test were used to assess normality. Continuous data are reported as mean ± standard deviation or as median (25th, 75th percentile), and categorical data as numbers and percentages. Student's t test, Oneway analysis of covariance (ANCOVA) and Mann Whitney U were used to compare continuous data. Chi-square and Fisher's exact test were used for comparisons of categorical data.
Cardiac function was compared in participants with normal (VO 2peak > 85% of predicted) verse reduced (VO 2peak < 85% of predicted) exercise capacity. ANCOVA was used to control for potential confounders to cardiac function created by this sub-grouping. Covariates included age at examination, BMI, heart rate (HR) and SBP. An additional supplementary analysis with ANCOVA, Kruskal-Wallis test and with Bonferroni correction was used to compare cardiac function between participants with mildly reduced (75%-85% of predicted-VO 2peak ), moderately reduced (<75% of predicted-VO 2peak ) and normal exercise capacity (>85% of predicted-VO 2peak ) (Table S1).
Pearson bivariate correlations determined the presence of linear relationships. Univariable and multivariable linear regressions were used to determine significant explanatory variables for VO 2peak (ml/kg/min). The multivariable analysis included a priori selected variables considered as central determinants for exercise capacity, and /or variables with p < 0.2 in the univariable analysis. All continuous variables were standardized and values presented with beta, confidence intervals and p value. Assumption testing included histograms, residual plots and assessment of multi-collinearity by Pearson correlations, tolerance and variance inflation factor (VIF). Considerations were made to avoid over-fitting. Pairwise omission was chosen to handle missing data. Receiver operating characteristics (ROC) were performed and areas under the curve (AUC) calculated to test the parameter's ability to identify patients with reduced exercise capacity.

| RESULTS
Two-hundred and ninety patients received allo-HSCT in the time frame specified for this study. Of these, 131 (45.2%) died prior to study start and two were excluded due to incomplete patient files ( Figure 1). One-hundred and fifty-seven were eligible for inclusion, of which 104 (66.2%) were examined with echocardiography, spirometry and blood tests, and 96 (61.1%) examined with CPET. Eight participants were excluded from with CPET due to musculoskeletal disorders (n = 3), significant systemic hypertension and reduced cardiac function (n = 2), suspected coronary artery disease or abnormalities (n = 2) and congenital aortic stenosis (n = 1). Table 1

| CPET and exercise capacity
The main findings from CPET are summarized in Table 2. All individuals were exercised to peak effort: Borg scale ≥18 and/or respiratory exchange ratio (RER) ≥1. 10

| Predictors of VO 2peak
VO 2peak (ml/kg/min) significantly correlated with BMI (R = À0.45,  Recently, we reported on pulmonary function in relation to cardio-respiratory fitness in this cohort. 20   Echocardiography performed at rest is a surrogate measure of actual myocardial function during exercise. However, multiple studies (mostly unrelated to cardiotoxic therapies) have shown relationships between functional capacity and echocardiographical measurements of systolic function, 11,12,31 diastolic function 32,33 and RV function. 34,35 In this cohort, we found multiple and overlapping phenotypes of cardiac dysfunction. Most common (44%) were cases of mild to moderate LVSD that were strongly associated with RVSD. 13  In similarity with the larger St. Jude Lifetime cohort by Ness et al.,

| N-terminal pro-brain-type natriuretic peptide
that examined long-term (≥10 years) adult survivors of childhood cancer, we also found GLS to be superior to 2D-LVEF (and also 3D-LVEF in our study) for predicting exercise capacity. 12 A possible explanation for 2D-LVEF inability to establish associations with VO 2peak may in part be due to geometric assumptions used in its calculation. The higher sensitivity of GLS to detect mild reductions due to cardiotoxicity, better reproducibility and lower variability compared to 2D or 3D-LVEF are other possible reasons. [6][7][8][9] It is also likely that the level of association between parameters of systolic function and exercise capacity is underestimated in our study due to exclusion of several patients with heart disease for CPET.
This study also included the biomarker NT-proBNP owing to its moderate ability in predicting VO 2peak in patients with heart failure. 36 Most instances of elevated NT-proBNP in this cohort were found in survivors with more pronounced cardiac dysfunction. It was notable that NT-proBNP was within normal limits in many participants with cardiac dysfunction by imaging, and was not associated with VO 2peak .
CPET is considered the gold standard for assessing functional capacity in patients with heart disease and the assessment of VO 2peak provides valuable prognostic information, including all-cause mortality. [3][4][5] In childhood survivors of cancer with reduced exercise capacity (VO 2peak < 85% of predicted) the hazard rate for death increases by approximately fourfold. 12 While, a finding of VO 2peak > 20 ml/kg/min in patients with heart failure is considered to correspond to better short-term prognosis. 37 To our knowledge, there are no other data documenting the prognostic impact of mild or moderate reduction in exercise capacity in long-term survivors of HSCT. However, the ability to identify cardiac dysfunction as a cause for reduced exercise capacity (even if mild) has obvious medical value and potential prognostic benefits.
A limitation with CPET is its inability to distinguish specific cardiac mechanisms for reduced oxygen uptake. Thus, identifying tools that explain exercise intolerance is important for clinical decision-making. This is especially relevant in survivors of allo-HSCT with multiple comorbidities and with uncertain origins of functional dyspnea. We recommend GLS in screening of allo-HSCT survivors to confirm cardiac dysfunction (irrespective of symptoms) and to provide explanation for reduced oxygen uptake. As demonstrated in this study, GLS when measured at rest had superior ability to predict VO 2peak compared with NT-proBNP and other tested echocardiographical parameters. Lower measurement variability for GLS is one possible explanation for this occurrence. Moreover, GLS has been shown to have valuable prognostic capacity. 6,7,10 Our study was not designed to address mortality, although based on our findings we are supportive. Finally, the use of GLS is endorsed by experts in cardio-oncology for early detection of cardiotoxicity, which assists in earlier therapeutic interventions to hinder progressive heart disease associated with reductions in LVEF. 28

FUNDING INFORMATION
The Norwegian Extra-foundation and the Norwegian Cancer foundation provided funding for this study. Open access publication is funded by Oslo University, Norway. No funding was received from the private industry.

DATA AVAILABILITY STATEMENT
The datasets generated and/or analyzed during the current study are not publicly available due restrictions set by Norwegian Regional Committee for Medical Research Ethics, but are available from the corresponding author on reasonable request.