Prognostic value of peak work rate indexed by left ventricular diameter

Left ventricular diameter (LVEDD) increases with systematic endurance training but also in various cardiac diseases. High exercise capacity associates with favorable outcomes. We hypothesized that peak work rate (Wpeak) indexed to LVEDD would carry prognostic information and aimed to evaluate the association between Wpeak/LVEDDrest and cardiovascular mortality. Wpeak/LVEDDrest (W/mm) was calculated in patients with an echocardiographic examination within 3 months of a maximal cycle ergometer exercise test. Low Wpeak/LVEDDrest was defined as a value below the sex- and age-specific 5th percentile among lower-risk subjects. The association with cardiovascular mortality was evaluated using Cox regression. In total, 3083 patients were included (8.0 [5.4–11.1] years of follow-up, 249 (8%) cardiovascular deaths). Wpeak/LVEDDrest (W/mm) was associated with cardiovascular mortality (adjusted hazard ratio (HR) 0.28 [0.22–0.36]), similar to Wpeak in % of predicted, with identical prognostic strength when adjusted for age and sex (C-statistics 0.87 for both). A combination of low Wpeak/LVEDDrest and low Wpeak was associated with a particularly poor prognosis (adjusted HR 6.4 [4.0–10.3]). Wpeak/LVEDDrest was associated with cardiovascular mortality but did not provide incremental prognostic value to Wpeak alone. The combination of a low Wpeak/LVEDDrest and low Wpeak was associated with a particularly poor prognosis.


Methods
We performed a retrospective analysis of consecutive patients aged 18 years or older who were referred for a clinical cycle ergometer exercise test at the department of Clinical Physiology at Kalmar County Hospital, Sweden between 31 May 2005 and 31 Oct 2016. The exercise stress test database has been described in detail elsewhere, and forms the basis for the Swedish national recommendations for grading of exercise capacity during standardized exercise stress testing 13,[22][23][24][25] . Within this database, patients who had performed an echocardiographic examination within 3 months from the date of the exercise stress test were included. Patients who did not reach a peak rating of perceived exertion (RPE) of 17/20, and patients with missing data on W peak or LVEDD rest were excluded. A flowchart of patient inclusion and exclusion is presented in Fig. 1.
To obtain survival status for each subject until study census date, the database was cross-linked with the mandatory Swedish Causes of Death Register (until 31 Dec 2019). Data on comorbidities, medications and hospital admission data were obtained through cross-linkage with the mandatory Swedish National Patient Register (until 31 Dec 2019) 26 . Definitions and diagnosis codes (International Classification of Diseases-10 (ICD-10)) are presented in Supplements (Table A). CV death was defined as death with an underlying cause within the ICD-10 chapter of CV disease (IX).
Exercise testing. Exercise testing was performed using a standardized protocol on an electrically braked computer-controlled, regularly calibrated cycle ergometer (Rodby Inc, Karlskoga, Sweden). A 12-lead electrocardiogram was recorded during rest, exercise and recovery. Systolic blood pressure was measured in the supine position before and after exercise, seated on the bicycle before exercise, and every 2 min during exercise. RPE was reported every 2 min during exercise. Work rate was started at 40-100 W (men) or 30-50W (women), depending on the expected exercise capacity, with an incremental increase of 10-20 W/min 27 . Patients were encouraged to exercise until exhaustion. The test was terminated at the patient's will or if predefined termination criteria were fulfilled (severe chest pain, ST depression ≥ 0.4 mV, decreasing systolic blood pressure or any malignant dysrhythmia). In order to take the effect of different work load increments on the achieved absolute W peak , into account, the achieved W peak was re-calculated, if necessary, to a standard protocol with an increment of 15 W/min (men) and 10 W/min (women) by the following formula 28 : females: W peak × (incremental workload used/10) 1/6 ); males: W peak × (incremental workload used/15) 1/6 . The re-calculated exercise capacity was then related to the Swedish reference material for standardized exercise stress testing (% of predicted W peak , which is adjusted for age, gender and height) 25 . ST-segment amplitudes were measured 60 ms following the J-point (ST60). Significant ST depression was defined as horizontal or down-sloping ST depression ≥ 0.1 mV in V5 during exercise or during the recovery phase. Heart rate recovery was defined as the difference in heart rate between the maximal heart rate and the heart rate 2 min after cessation of exercise 15,29 . If a patient had performed more than one test, only the most recent test was included.
Echocardiography. Two-dimensional transthoracic resting echocardiography was performed using commercially available echocardiographic equipment. At end-diastole, LVEDD rest (mm), septal and posterior wall thickness (IVS and PWT, mm) were obtained either from M-mode or 2D images in the parasternal long axis view. LV mass was calculated according to the Cube formula (LV mass (g) = 0.8 × 1.04 ([IVS + LVEDD + PWT] 3 − LVEDD 3 ) + 0.6) and presented after adjustment to body-surface (BSA) 30 . Left ventricular ejection fraction (LVEF) was reported either based on the Simpson biplane method, from M-mode data (Teichholtz formula), or by visual estimation 31 .
Aortic, mitral or tricuspid regurgitation were graded as none, mild, moderate or severe. Moderate aortic stenosis was defined as either an aortic valve (AV) maximal velocity by continuous Doppler ≥ 3.1-4.0 m/s or an AV mean gradient of 20-40 mm Hg, while severe aortic stenosis corresponded to an AV maximal velocity ≥ 4.0 m/s or an AV mean gradient ≥ 40 mm Hg. Early (E) and late (A) diastolic velocity over the mitral valve were measured using pulsed Doppler and the E/A ratio was calculated. Pulsed tissue Doppler imaging with a 2-mm sample volume placed in the myocardium at the septal and lateral mitral annulus (apical four-chamber www.nature.com/scientificreports/ view) was used to determine the average early diastolic myocardial velocity (e'), in order to calculate the E/e' ratio. W peak /LVEDD rest (W/mm) was calculated as the peak work rate at exercise stress testing divided by the end-diastolic LV diameter during resting echocardiography.
Statistical analysis. Continuous variables were described using mean and standard deviation (SD). Comparisons of group means were performed using Student's t test. Differences between groups were assessed using the χ 2 test. The correlation between W peak and LVEDD was analyzed using Pearson's r and visualized using scatter plots, separately for all patients, lower-risk subjects, patients with moderate/severe left-sided valvular regurgitation, and for patients with heart failure. The association between W peak /LVEDD rest and CV mortality, and the association between W peak in % of predicted, was analyzed using Cox proportional hazard regression models. Models were evaluated unadjusted; adjusted for age and sex; and adjusted for age, sex, peak systolic blood pressure, presence of ST depression, heart rate recovery, peak heart rate, LVEF, E/e' , heart failure, hypertension, previous myocardial infarction, diabetes mellitus, hyperlipidemia, and peripheral arterial disease. The choice of confounding variables was based on directed acyclic graphs and previous literature knowledge. The assumption of proportional hazards was confirmed using Schoenfeld's residuals. Results are presented as hazard ratios (HR) with 95% confidence intervals (CI) and C statistics for the continuous W peak /LVEDD rest and W peak .
We also present HR for combinations of low/normal (W peak /LVEDD rest and low/normal W peak ). Since no reference values for W peak /LVEDD rest exist, an approximate age-and sex-specific lower limit of normal (LLN) for W peak /LVEDD rest was defined among a subgroup of lower-risk subjects in this cohort, including only nonobese (BMI < 30 kg/m 2 ) subjects with normal LVEF (≥ 55%), absence of moderate/severe valvular heart disease, without CV medications or known CV, renal, respiratory, or malignant disease. A low W peak was defined as W peak in % of predicted below the LLN according to the validated, Swedish reference material for exercise capacity 13,25 .
Natural cubic spline modelling was used to characterize the risk associated with W peak /LVEDD rest and W peak as a continuum, using four knots placed at the 5th, 25th, 75th and 95th percentiles.
Statistical significance was defined as a two-tailed p-value < 0.05. Statistical analysis was performed using R v. Permissions. The manuscript does not contain any reproduced material from other sources.

Results
A total of 3083 patients (mean age 60 ± 16 years, 55% male) were included. During a median follow-up of 8.0 [IQR 5.4-11.1] years, 592 (19%) patients died of whom 249 (8%) died due to a CV cause. Baseline characteristics are presented in Table 1. An echocardiographic examination was performed within 0 [0-1] days from the exercise stress test.
The mean W peak /LVEDD rest was 3.1 ± 1.2 W/mm in the whole study group. Males had higher W peak /LVEDD rest compared to females (3.6 ± 1.2 W/mm vs. 2.6 ± 0.8 W/mm, p < 0.001). W peak /LVEDD rest was lower in older age groups among both sexes (Supplemental Fig. A).
Mean W peak /LVEDD rest in HF patients was 2.6 ± 1.1 W/mm, 2.6 ± 0.9 W/mm in patients with at least moderate mitral or aortic regurgitation, and 3.3 ± 1.1 W/mm in patients with neither HF nor mitral or aortic regurgitation (p < 0.001 for both comparisons).
A total of 460 patients (15%) met the definition as lower-risk subjects. Among these, W peak was moderately correlated with LVEDD rest (r = 0.61, p < 0.001) (Fig. 2), while there was no clinically significant correlation between W peak and LVEDD rest among patients with HF or valvular heart disease (r = 0.06, p = 0.31 and r = 0.22, p < 0.001 respectively). A low W peak /LVEDD rest , defined as less than age-and sex specific 5th percentile of the lower-risk subgroup ( Table 2), was found in 1,362 patients (44.2%). Compared to subjects with W peak /LVEDD rest above this threshold, patients with a low W peak /LVEDD rest had a higher prevalence of low LVEF < 50% (14.2% vs. 3.5%, p < 0.001), while they had greater E/e' and LV mass (12 vs. 9, p < 0.001; 109 g/m 2 vs. 92 g/m 2 , p < 0.001 respectively), Table 1.
Higher W peak /LVEDD rest was associated with reduced CV mortality (unadjusted HR: 0.26 [0.22-0.31] per 1 W/mm), Table 3, Fig. 4. C-statistics were numerically higher for W peak /LVEDD rest than for W peak in % of pre- The relative risk of CV mortality increased exponentially with lower W peak /LVEDD rest , as well as with lower W peak , Fig. 3. The combination of low W peak /LVEDD rest and a low W peak was associated with a particularly poor prognosis (adjusted HR 6.4 [4.0-10.3]) in reference to patients with normal W peak /LVEDD rest and normal W peak ), Fig. 4. In reference to those with a low W peak /LVEDD rest but a normal W peak , the risk was three folded increased for those with both low W peak /LVEDD rest and low W peak (

Discussion
In a large consecutive cohort of patients undergoing clinical cycle ergometer exercise testing and an echocardiogram within 3 months, W peak /LVEDD rest was a strong predictor of CV mortality, similar to W peak in % of predicted, with identical prognostic strength when adjusted for age and sex. We hypothesized that patients with a low exercise capacity and an enlarged LV would be at higher risk of cardiovascular death than patients with a low exercise capacity but a normally sized LV. While we did not find W peak / LVEDD rest to provide incremental value to W peak alone, we found that a low W peak /LVEDD rest in combination with a low W peak was associated with a particularly poor prognosis. A combination of a low W peak /LVEDD rest and a low W peak increased the risk of cardiovascular death by more than 600% (adjusted HR 6.4 [4.0-10.3])), in reference to those with normal W peak /LVEDD rest and a low W peak . The risk of cardiovascular death increased exponentially with lower W peak /LVEDD rest . Therefore, it is possible that the W peak /LVEDD rest could have a potential value in  www.nature.com/scientificreports/ Table 3. Associations between peak work rate to left-ventricular end-diastolic diameter at rest (W peak / LVEDD rest ) and W peak (% of predicted) and LVEDD rest , respectively, and cardiovascular mortality (n = 3083 (249 events)). HR hazard ratio, LVEDD left ventricular end-diastolic diameter, W Watt, W peak peak work rate. *Adjusted for age, sex, peak systolic blood pressure, ST depression, heart rate recovery, peak heart rate, left ventricular ejection fraction, E/e, heart failure, hypertension, myocardial infarction, diabetes mellitus, hyperlipidemia, and peripheral arterial disease.  Figure 3. Impact of decreasing W peak /LVEDD rest and W peak alone on the risk of cardiovascular death. The hazard ratios (95% confidence intervals) were calculated using Cox regression and modelled with natural cubic splines with four knots (percentiles: 5th, 25th, 75th, 95th) and presented as unadjusted estimates (left panels) and adjusted for age and sex (right panels). W peak /LVEDD rest showed a substantial risk increase with low values both for adjusted and unadjusted estimates. www.nature.com/scientificreports/ risk stratification of patients with heart failure in settings where more advanced markers for risk assessment are unavailable, e.g. cardiovascular magnetic resonance imaging or cardiopulmonary exercise stress testing with breathing-gas analysis 33,34 . When describing the relation between exercise capacity and LV size, neither W peak nor LVEDD were indexed to body size or age. When determining whether exercise capacity is reduced or not, W peak should be related to anthropometric data (including sex and age) 13,18 . Reference values for exercise capacity using standardized cycle ergometer exercise test in Sweden has been published, in which age, sex and height, but not weight, are included in the regression equation 25 . In this study, we aimed to test the hypothesis that if exercise capacity was not increased in parallel to cardiac dimensions, the risk of cardiovascular death would increase. Since both exercise capacity and LVEDD are expected to be higher in larger individuals, this hypothesis does not require for any of the measures to be adjusted to body size. Previous studies have applied a similar strategy, i.e. assessing absolute values for peak VO2 and total heart volume or left ventricular dimension instead of indexing to body size 2,35 .
W peak correlated with LVEDD rest in the subgroup of lower-risk subjects, similar to previous reports in healthy subjects 1,3,4,36 . It has been shown previously that intensive endurance training leads to an increase in LV-volume in healthy subjects, as a cardiac morphological adaptation 4 , and Meyer et al. showed that patients with a low peak work rate tended to have smaller left ventricles. Our study showed no meaningful correlation between LV size and peak work rate in patients with HF or valvular heart disease. This is not unexpected since the underlying cause of LV remodeling in response to endurance training is different than, for example, after myocardial infarction or with valvular heart disease. In contrast, among lower risk subjects, the association between LV size and W peak was stronger. In that way, an LV dilatation which is disproportionate to W peak is a sign of LV disease, in our study represented by a low W peak /LVEDD rest 2 . Although this could not be elucidated in the present study, the potential role of W peak /LVEDD rest in the discrimination of LV remodeling in response to exercise, i.e. athlete's heart, and cardiac disease could warrant future studies.
Limitations. Firstly, as we included patients with a clinical referral to exercise stress testing and echocardiography, there is a selection bias, limiting broad generalization of our findings. Secondly, LV volumes were not available, and may have provided better assessment of LV dilatation than LVEDD. It is also possible that including measures of both ventricles and atria may be a better way to index the peak work rate to cardiac size, since physiological cardiac adaptation generally affects all four cardiac chambers 2 .
Thirdly, echocardiographic data were obtained from clinical records and not by a standardized, study-specific protocol.
Since this is a retrospective analysis, we have no data on reproducibility of the LVEDD values. . Time-to-event analysis for the combination normal/low W peak /LVEDD rest and normal/low W peak 3083 patients experiencing 249 events (cardiovascular death) over a median of 8.0 [IQR: 5.5, 11.1] years. A low W peak / LVEDD rest as well as a low W peak was defined as a value below the sex-and age-specific 5th percentile of lowerrisk subjects.

Conclusions
W peak /LVEDD rest was associated with cardiovascular mortality but did not provide incremental prognostic value to W peak alone. However, the combination of having a low W peak /LVEDD rest and low W peak was associated with a particularly poor prognosis.

Data availability
The data that support the findings of this study are available upon reasonable request to the corresponding author.