Aerobic Training Protects Cardiac Function During Advancing Age: A Meta-Analysis of Four Decades of Controlled Studies

Background In contrast to younger athletes, there is comparatively less literature examining cardiac structure and function in older athletes. However, a progressive accumulation of studies during the past four decades offers a body of literature worthy of systematic scrutiny. Objectives We conducted a systematic review, meta-analysis and meta-regression of controlled echocardiography studies comparing left ventricular (LV) structure and function in aerobically trained older athletes (> 45 years) with age-matched untrained controls, in addition to investigating the influence of chronological age. Methods Electronic databases were searched from inception to January 2018 before conducting a random-effects meta-analysis to calculate pooled differences in means, effect size and 95% confidence intervals (CIs). Study heterogeneity was reported using Cochran’s Q and I2 statistic. Results Overall, 32 studies (644 athletes; 582 controls) were included. Athletes had greater LV end-diastolic diameter (3.65 mm, 95% CI 2.66–4.64), interventricular septal thickness (1.23 mm, 95% CI 0.85–1.60), posterior wall thickness (1.20 mm, 95% CI 0.83–1.56), LV mass (72 g, 95% CI 46–98), LV mass index (28.17 g·m2, 95% CI 19.84–36.49) and stroke volume (13.59 mL, 95% CI 7.20–19.98) (all p < 0.01). Athletes had superior global diastolic function [ratio of early (E) to late (A) mitral inflow velocity (E/A) 0.18, 95% CI 0.13–0.24, p < 0.01; ratio of early (e′) to late (a′) diastolic annular tissue velocity (e′/a′) 0.23, 95% CI 0.06–0.40, p = 0.01], lower A (−8.20 cm·s−1, 95% CI −11.90 to −4.51, p < 0.01) and a′ (−0.72 cm·s−1, 95% CI −1.31 to −0.12, p = 0.02), and more rapid e′ (0.96 cm·s−1, 95% CI 0.05–1.86, p = 0.04). Meta-regression for chronological age identified that athlete–control differences, in the main, are maintained during advancing age. Conclusions Athletic older men have larger cardiac dimensions and enjoy more favourable cardiac function than healthy, non-athletic counterparts. Notably, the athlete groups maintain these effects during chronological ageing. Electronic supplementary material The online version of this article (10.1007/s40279-018-1004-3) contains supplementary material, which is available to authorized users.


Introduction
It is well established that a lifestyle consisting of regular physical exertion is associated with reduced cardiovascular risk and all-cause mortality across the age spectrum [1,2]. Despite this, epidemiological studies consistently identify older adults as the least physically active demographic where few achieve the recommended levels of physical activity required to accrue health benefits [3]. Studies of ageing athletes observe relatively high levels of cardiovascular reserve (maximal aerobic capacity) compared with their sedentary counterparts [4]. In this respect, the 'masters athlete' model offers a unique non-pharmacological model to differentiate the inexorable from the preventable mechanisms of cardiovascular ageing. In studies of cardiovascular function, endurance-trained masters athletes have superior functional capacity and cardiovascular reserve to their sedentary peers, and are comparable in these respects to non-athletes who are years (and often decades) their junior. As this is not the focus of the present review, we direct readers to the eminent works from Professor Benjamin Levine (University of Texas) and Professor Douglas R. Seals (University of Colorado) [5,6].
A major limitation of this area, however, is that randomised controlled trials of life-long exercise are highly unlikely, and therefore researchers must rely on evidence from animal models and controlled observational studies to further our understanding of the effects of age on cardiac structure and function [5,[7][8][9]. Indeed, the pleiotropic benefits of aerobic endurance exercise have been acknowledged as a potential non-pharmacological mitigant to deterioration of left ventricular (LV) diastolic function [10]. Although this body of evidence is encouraging, it is often hampered by small sample sizes in controlled observational studies which suffer from low statistical power, methodological inconsistencies and untested external validity. In contrast to a wealth of data in younger cohorts < 45 years, including several meta-analyses of LV, right ventricular (RV) and left atrial (LA) structure and function in both male [11][12][13][14][15][16][17] and female [14,17,18] athletes, there have been narrative reviews [10,[19][20][21] but no systematic synthesis of evidence in older athletes relative to matched, untrained controls. Further, in the absence of prospective investigations of multiple decades in duration, quantifying the influence of chronological age on exerciserelated morphological and functional adaptations must be realised through cross-sectional studies, across the age spectrum from middle to older age (i.e. > 45 years).
In light of this, we undertook a systematic review and meta-analysis of studies using echocardiography to (1) compare LV morphology and systolic and diastolic function in older athletes versus age-matched sedentary but otherwise healthy controls and (2) employ meta-regression to explore the influence of chronological age on cardiac structure and function in the ageing athlete.

Methods
The systematic search processes, evaluation, analysis, and reporting were conducted and presented in accordance with the Preferred Reporting Items for Systematic review and Meta-Analyses (PRISMA) guidelines for reporting systematic reviews and meta-analyses [22].

Information Sources and Search Strategy
An electronic database search was designed by two authors (AB and NS), who conducted an independent literature search of PubMed (title and abstract), MEDLINE (title) and ScienceDirect (title, abstract and keywords) for published, English-language journal articles from inception to 11 January 2018 (see Electronic Supplementary Material Appendix S1 for the PubMed database search).

Inclusion Criteria
Studies were eligible for quantitative analysis when meeting the following criteria: (1) study participants were male; (2) mean participant age ≥ 45 years; (3) aged-matched control group; (4) athlete groups were aerobic/endurance trained; (5) control groups were untrained; (6) participants were reported as free from cardiovascular diseases; (7) the study was observational in design and data were recorded at a single time point (including recruitment data from intervention studies); (8) studies used echocardiography; and (9) studies assessed cardiac strain using two-dimensional speckle tracking echocardiography (STE).
We elected to include only endurance athletes participating in predominantly dynamic aerobic activity (for example, distance running, cycling, and rowing). Where study information was unclear, corresponding authors were contacted by email.
Multiple segment reports for mitral annular tissue velocities (septal, lateral wall, inferior wall, anterior wall) were combined to obtain a global value. We included studies employing either pulsed-wave Doppler or colour Doppler techniques to obtain tissue velocities. Data for RWT included both descriptions of RWT and thickness/radius ratio. We elected to only include studies that scaled LVM to body surface area (BSA) in line with current echocardiographic recommendations [23]. GLS was deemed as a global value from the average of multiple segments, and E/e′ was considered to be an estimate of LV filling pressure [24].
When unsuccessful attempts were made to contact authors, data were extracted from study figures. Study means ± standard deviation (SD) were recorded for all variables; where studies reported the standard error of the mean (SEM), we applied a manual conversion using the formula SD = SEM × √N, where N is the number of participants. Likewise, when median and range were reported, a manual conversion was applied to convert data into mean and SD, in accordance with the study sample size [25]. For each study, the mean age of athlete and control groups were averaged to obtain a pooled mean. Study quality was assessed for each individual study using a 17-point checklist (Electronic Supplementary Material Table S1) adapted for specificity for this meta-analysis from a previously published checklist used in a similar study of young athletes [13].

Statistical Analyses
Meta-analyses were executed using Comprehensive Meta-Analysis (Biostat, V 2.2.064, Englewood, NJ, USA). Pooled data using a random-effects model were used to investigate athlete-control differences. Differences in means were calculated for each individual study, and a summary of overall difference in means recorded for all variables. Differences in means in a positive direction indicated greater magnitude of LV structure or function in athletes, with negative direction favouring a greater magnitude in controls. Heterogeneity was reported using Cochran's Q and I 2 statistic and classed as either low, moderate, or high at 25%, 50%, and 75%, respectively [26]. Using pooled athlete and control age (continuous moderator variable), we conducted randomeffects (method of moments) meta-regression analysis to examine the relationship between differences in means with chronological age. Meta-regression analysis was performed where there were ten or more studies [27]. Publication bias was addressed using Egger's regression intercept [28] to test for asymmetry and interpreted conservatively [27]. Statistical significance was granted at p ≤ 0.05. Figure 1 illustrates the systematic filtration process. The electronic database search resulted in 597 records, which were exported to referencing software (Zotero, Fairfax, VA, USA) to manage the systematic process.

Search Outcome
Thirty-two studies including 1226 participants [644 athletes (59 ± 8 years) and 582 controls (59 ± 8 years); mean age range 46-74 years] met the inclusion criteria for quantitative analyses [7,. This allowed for the following comparisons [number of studies presented in parentheses; study names for each comparison can be found in sup Summarised participant anthropometrics, systolic and diastolic blood pressure and heart rate are presented in Table 1. Characteristics of athlete training regimes and control activity levels are reported in Table 2. LV, RV and LA structure and systolic and diastolic function data for control and athlete groups are presented in Tables 3, 4, and 5, respectively. Table 6 describes the meta-analyses data for athlete-control comparisons including between-study heterogeneity. Electronic Supplementary Material Figs. S1-S36 present forest plots of athlete-control comparisons and meta-regression figures of the athlete-control differences moderated for chronological age.

Cardiac Structure
IVS, PWT, LVEDD, LVEDV, LVM and LVMi were greater in athletes compared with controls, while RWT, RVEDD and LAD did not differ. Significant heterogeneity was observed for all parameters, with inconsistency considered moderate for LVEDD, yet high for IVS, PWT, RWT, LVM, LVMi, LVEDV RVEDD and LAD.

Left Ventricular Systolic Function
EF, FS, s′ and GLS were not different between athletes and controls, whereas SV was greater in athletes. Betweenstudy heterogeneity was significant in all cases and the inconsistency considered moderate in FS and GLS, yet high for EF, s′ and SV.

Left Ventricular Diastolic Function
Pooled analysis of studies measuring mitral inflow velocity revealed no difference in E between athletes and controls, albeit A was significantly lower in athletes. Accordingly, E/A was greater in athletes compared with controls. Tissue  Cottini et al. [34] Control ( Olsen et al. [56] Control (11) 66 Miki et al. [51] Control (14) 49.0 ± 7.6 --- Maufrais et al. [49] Control (20) 58.0 ± 7.0 177 ± 6 75.5 ± 11.8 - Maessen et al. [47] Control (13) 55 velocity analyses showed higher e′ in athletes, whereas athletes showed significantly reduced a′ but greater e′/a′ compared with controls. E/e′ was not different between athletes and controls. Between-study heterogeneity was significant for all parameters of diastolic parameters besides E/e′. Inconsistency was low for E/A and E/e′, moderate for E, A and a′, and high for e′ and e′/a′.

Meta-Regression(s)
The difference between athletes and controls was maintained with chronological age for all available variables, with the exception of LVEDD, LVMi and A (Table 7). There was a significant inverse but opposite relationship in LVEDD, LVMi and A between athletes and controls that continued with chronological age.

Publication Bias
Egger's regression revealed that LVMi was the only variable which demonstrated significant bias (Electronic Supplementary Material Table S2).

Discussion
This first systematic pooling of controlled echocardiographic evidence of cardiac structure and function in aerobically (endurance) trained older athletes compared with age-matched untrained controls reveals that (1) endurance-trained older athletes have superior diastolic function compared with untrained counterparts, which is sustained regardless of advancing age; (2) mean LVM of older endurance athletes is greater (mean difference 72 g) than that of controls, as evidenced by greater wall thickness (mean ~ 1.2 mm) and chamber diameter (mean difference 3.7 mm); (3) despite modest differences in global systolic function between groups, older athletes have a larger SV (equivalent to mean difference ~ 1 L·min −1 greater cardiac output at rest) than controls, and this difference is sustained with chronological ageing.

Left Ventricular Diastolic Function
Older athletes enjoy more favourable global diastolic function which is driven by a reduced reliance on late diastolic filling and evident in both haemodynamic and tissue assessment. Greater E/A in aerobic athletes both contrasts [12] and concurs [13] with previous meta-analyses in younger cohorts. The E/A difference between groups shown here (mean difference 0.18) compares well with that reported for younger individuals (mean difference 0.2) by Utomi et al. [13], suggesting that the difference in global diastolic function between aerobically trained athletes and untrained controls is age independent. This is endorsed by the finding of the present meta-regression. Collectively, this supports the tenet that global diastolic function assessed by mitral inflow is preserved in athletes (versus controls) from ~ 18 to 74 years of age. The improved global diastolic function mediated by reductions in A, without concomitant changes in E, accords with a large investigation of young, Olympic calibre athletes [60], highlighting a similar adaptation in athletes of all ages compared with untrained controls. Moreover, meta-regression analysis observed that the difference in A between athletes and controls becomes more exaggerated with advancing age, which suggests an incremental reliance on atrial contraction to support LV filling in untrained controls.
Greater e′ in older trained adults agrees with findings from Utomi et al. [13] in younger athletes. In our meta-analysis, athletes demonstrated lower a′, and when considered with the observed difference in early diastole, a greater e′/a′ was observed in athletes than in controls. These observations support the precept that continuation of a high volume of aerobic exercise into advanced age has pleiotropic effects on cardiac function.
Similarly, RVEDD did not differ between older athletes and controls, which agrees with previous work in younger endurance athletes. Yet, two of the three studies in this metaanalysis observed larger RVEDD in athletes than in controls, highlighting a biventricular enlargement [30,35] but significant heterogeneity between studies. This could be explained by age, as the participants in the studies showing RV dilation were younger than those in the study showing comparable diameter between athletes and controls [37]. Given the observed meta-regression of reduced mean difference LVEDD between athletes and controls with chronological age, it may be possible for a similar observation regarding the right-sided heart; however, more data are required before a sufficiently powered meta-regression can be conducted.

Left Ventricular Systolic Function
Conventional measures of LV systolic function, EF and FS were comparable between athletes and controls, as observed in younger cohorts [12,13], which indicates preservation of resting EF and FS in older athletes. SV was greater in athletes than in controls and comparable with the literature for younger athletes [13]. In the absence of changes to EF, greater SV in athletes is likely a reflection of larger LVEDV and LVEDD [66], which is notably preserved despite increasing age.
Speckle tracking-derived GLS is a more sensitive marker of LV systolic (dys)function than conventionally measured EF [67]. However, GLS appeared to be unaltered between older aerobically trained athletes and controls, which agrees with the findings of a recent systematic synthesis of data for younger athletes [15]. There is emerging evidence that GLS is reduced in disease states [68][69][70] and associated with poor cardiac outcomes [71]. Studies of younger [72][73][74] and older individuals [75] have reported small increases in GLS after aerobic training programmes, and these are considered to be positive adaptations of the athlete's heart [76]. However, with only five studies included in this meta-analysis, the small volume of work in older athletes requires expansion before more accurate estimations can be made, especially considering recent documentation of preserved GLS in lifelong exercisers, albeit in a mixed sex cohort [77].

Available Studies
Within the present systematic search there was a single study of female athletes [78]. Due to this near absence of literature in ageing female athletes, the present findings should only be generalised to ageing men. Similarly, there were only Maessen et al. [47] Control (13) Endurance athletes (18) ------9.0 ± 1.4 9.3 ± 1.9 ↔ --Maufrais et al. [49] Control (20) Cyclists (22) ----   [42] Control (12) Cyclists (12) 54  [45] Control (9) Cyclists, triathletes, speed-skaters (12) 73  [50] Control ( [41] Control ( [35,79] in ageing men, which precluded comparisons between the two most common exercise modes. In particular, more data are needed to fully characterise LA size in masters athletes, including diameter and volume, since varying degrees of lifelong training hours were associated with LA volume but not dimension [80]. The preponderance of studies of the LV and paucity of studies reporting on the RV limited our inclusion of RV indices, and therefore, a greater focus on the right side of the heart is warranted in older athletes. Future work should also include STE assessment of both left and right sides of the heart, to enable further insight into intrinsic cardiac mechanics.
Reporting of athlete training was inconsistent between studies, which prevented additional meta-regression analyses of training years, volume and intensity to further elucidate the between-study heterogeneity. Therefore, more robust documentation with quantitative means ± SDs is required before the influence of training regimes can be documented within a narrower age range, since cardiac structure and function have been reported to be exercise dose dependent [81]. Many studies did not state the specific sport performed by the athletes, or enrolled endurance athletes from different sports with varying magnitudes of static and dynamic loading. To allow subgroup analysis on the type of sport in future meta-analysis, we suggest studies report the sports of their athletes.

Present Meta-Analysis
This meta-analysis has notable limitations. Firstly, we were confined to analysing male participants. Further, we did not include other measures of body size-adjusted LV morphology (IVS, PWT, LVEDD). This was due to few data and inconsistent scaling within the literature. Of the included studies for LVMi, five used either height 2.7 or fat-free mass to index LVM, which prevented separate meta-analysis and regression using these parameters for indexing other than BSA. It is noteworthy that two studies [34,51] did not disclose the sex of participants. To resolve this, we employed an educated assumption of male participants because average LVEDD and LVEDV greatly exceeded the normal range in females [34]. This was further endorsed when average group LVMi data scaled to BSA were at the upper end of the range in men [23,51] and vastly exceeded the expected range in females. We acknowledge that the data presented in this meta-analysis are derived from cross-sectional studies. However, there exists a paucity of prospective studies of sufficient duration to adequately quantify the influence of chronological age on the observed athlete-control differences, in addition to having relatively small sample sizes and mixed sexes. Indeed, 1 year of endurance training in older adults (> 65 years) did not alter LV stiffness and compliance [9], whereas recently, Howden et al. [82] observed reduced LV stiffness in middle-aged adults following 2 years of high-intensity training. While these data provide invaluable knowledge pertaining to the optimal stage of life to initiate exercise for offsetting the ageing process, the present findings document the benefit of exercise in those already trained across the age spectrum.

Conclusions
Pooled information from controlled echocardiography studies demonstrates that older athletes have superior global diastolic function compared with controls because of a reduced reliance on atrial contraction to LV filling in mitral inflow velocity and both increased early and reduced late diastolic tissue velocities. Furthermore, older athletes have notable differences in cardiac structure (wall thickness, cavity size and LV volume) with greater relative and absolute LVM. Despite unremarkable differences in LV systolic function, SV is markedly greater in older athletes. Notably, the present data also identify that masters athletes maintain these functional effects during chronological ageing.

Compliance with Ethical Standards
Funding No sources of funding were used to assist in the preparation of this article.