Article Text

Download PDFPDF

Systematic review and meta-analysis of training mode, imaging modality and body size influences on the morphology and function of the male athlete's heart
  1. Victor Utomi1,
  2. David Oxborough1,
  3. Greg P Whyte1,
  4. John Somauroo1,2,
  5. Sanjay Sharma3,
  6. Rob Shave4,
  7. Greg Atkinson5,
  8. Keith George1
  1. 1Cardiorespiratory research group, Research Institute for Sport and Exercise Sciences, Liverpool John Moores University, Liverpool, UK
  2. 2Cardiology Department, Countess of Chester Hospital, Chester, UK
  3. 3Cardiology Department, St George's Medical School, Tooting, London, UK
  4. 4Physiology and Health, School of Sport, Cardiff Metropolitan University, Cardiff, UK
  5. 5Health and Social Care Institute, Teesside University, Middlesborough, UK
  1. Correspondence to Dr Victor Utomi, Cardiorespiratory research group, Research Institute for Sport and Exercise Sciences, Liverpool John Moores University, Tom Reilly Building, Byrom Street, Liverpool L3 3AF, UK; v.s.utomi{at}


Context The athlete's heart (AH) remains a popular topic of study. Controversy related to training-specific cardiac adaptation in male athletes, and continuing developments in imaging technology and scaling prompted this systematic review and meta-analysis.

Objective To provide new insight in relation to: 1) cardiac adaptation to divergent training patterns in male athletes, 2) a developing research database using cardiac magnetic resonance (CMR) in athletes; 3) functional data derived from tissue-Doppler analysis as well as right ventricular (RV) and left atrial (LA) measurements in athletes; and 4) an awareness of the impact of body size on cardiac dimensions.

Study design Systematic review and meta-analysis of prospective trials. Data extraction performed by two researchers.

Data sources Pub Med, Medline, Scopus and ISI Web of knowledge scholarly data base.

Study selection Prospective studies were included if they were echocardiographic or CMR trials of elite young male athletes, with clear indication of type of sports and passed a quality criteria checklist.

Results All left ventricular (LV) structural parameters were higher in athletes than in controls. Only LV end-diastolic diameter and volume were higher in endurance athletes than in resistance athletes: 54.8 mm (95% CI 54.1 to 55.6) vs 52.4 mm (95% CI 51.2 to 53.6); p<0.001 and 171 ml (95% CI 157 to 185) vs 131 ml (95% CI 120 to 142); p<0.001, respectively. RV end-diastolic volume, mass and LA diameter were higher in endurance athletes than controls. LV end-diastolic volume was larger when CMR was used rather than echocardiography: 178 ml (95% CI Q7 162 to 194) vs 135 ml (95% CI 128 to 142); p<0.001. Meta-analysis regression models demonstrated positive and significant associations between body surface area (BSA) and LV mass, RV mass and LA diameter.

Conclusions Morphological features of the male AH were noted in both athlete groups. A training-specific pattern of concentric hypertrophy was not discerned in resistance athletes. Both imaging mode and BSA can have a significant impact on the interpretation of AH data.

View Full Text

Statistics from


The athlete's heart (AH) has been extensively described in male athletes.1 A cornerstone of this phenomenon has been the concept that cardiac structural adaptation follows a dichotomous course of eccentric hypertrophy (balanced increase in chamber and wall dimensions) with endurance training versus concentric hypertrophy (disproportionate increase in wall thickness)2 with resistance training.3–5 It is hypothesised that these adaptations reflect differential haemodynamic loading during acute training.6 The 26th Bethesda Conference adopted training-specific changes in cardiac morphology as a basis for its recommendations for cardiovascular screening,7 and this ‘hypothesis’ has been broadly reported in textbooks and teaching.

Despite this pervasive ‘knowledge’, contradictory evidence exists,8 ,9 mainly reflecting a lack of concentric hypertrophy in resistance-trained athletes.10 Haykowksy et al,10 suggested that the stimulus for concentric remodelling, a haemodynamic pressure overload on the left ventricle, may not occur during heavy resistance training due to a simultaneous Valsalva manoeuvre.

Allied to this continuing controversy, two specific technical issues warrant evaluation. First, only echocardiographic studies of the AH were included in a previous meta-analysis.5 Cardiac magnetic resonance (CMR) has now become the ‘gold standard’ for cardiac structural assessment. Clinically significant differences between echocardiography and CMR have been reported.11 While eccentric hypertrophy has been reported when CMR is used in endurance-trained participants,12 no concentric adaptation has been noted in resistance-trained athletes.13 Developments in CMR and echocardiography have also resulted in new regional functional indices (ie, tissue-Doppler analysis) and greater access to morphological and functional data from the right ventricle and left atrium. A second concern is the impact of body size on cardiac structure. Pluim et al,5 reported absolute cardiac dimensions in their between-group comparisons. Although informative, their data do not take into account potential differences in body size and composition between athletes and sedentary controls that might alter the interpretation of AH studies.14–17

Consequently, the aim of this systematic review and meta-analysis was to provide new insight in relation to (1) cardiac adaptation to divergent training patterns in male athletes; (2) a developing research database using CMR in athletes; (3) functional data derived from tissue-Doppler analysis as well as right ventricle and left atrium measurements in athletes and (4) an awareness of the effect of body size on cardiac dimensions. This study applied a quality-criteria guided approach to study selection, common in current meta-analyses, but not clearly applied in previous AH meta-analyses.5 ,18


Search criteria and processes

Our initial aim was to identify all echocardiographic and CMR studies examining the AH in male populations, published between 1975 and 2012, using a number of electronic search engines (eg, Pub Med, Medline, Scopus and ISI Web of Knowledge scholarly database). Relevant MeSH subject terms and keywords pertaining to the structure and function of the AH and Boolean operators were used in two separate searches for echocardiographic and CMR studies.

The following search strings were employed:

  • ‘Cardi$ OR Ventric$ OR Atria$ AND Athlet$ AND Echocardio$’

  • ‘Cardi$ OR Ventric$ OR Atria$ AND Athlet$ AND Magnetic resonance imag$’

  • ‘Cardi$ OR Ventric$ OR Atria$ AND Athlet$ AND MRI$’.

In addition to date limits, we confined the search to human studies and those with an English language abstract. Finally, this initial search was extended by a thorough cross-reference to reference lists from previous reviews and meta-analyses to find studies not cited in electronic databases.

The search process identified 378 echocardiography and 41 MRI records for potential inclusion. The filtration process from this point is detailed in figure 1. The first stage of the filtration process reviewed titles and then abstracts. This process was completed independently by two authors (VU, KG) who compared decision-making and discussed disagreements. Inclusion criteria were (1) male subjects, (2) age 18–45 years, (3) elite athletes with international or national level of participation for at least 2 years, (4) clear ability to define athletes as endurance or resistance focused, (5) healthy subjects with no history of cardiovascular disease and (6) an age-matched sedentary control group with low documented levels of physical activity. We excluded review papers and trials on children, veteran athletes, female athletes, mixed-sex athlete groups, patient groups, steroid users and sporting groups with a cross-training focus (eg, boxing).

Figure 1

Flow diagram of literature filtration process use.

Sixty-six records were excluded at the title level largely owing to a non-athlete focus to the studies. Examination of the echocardiography records at the abstract filtration level identified and excluded 127 studies, of which 101 papers had no control group. A further nine studies were rejected as athletes were not deemed to be elite. Thirteen CMR studies were excluded owing to lack of control group (n=5); non-elite athletes (n=2); veteran athletes (n=2); no cardiac data (n=2) as well as case studies and reviews (n=2).

All remaining records were obtained for the full record appraisal level. At this stage 116 records were excluded on assessment of the full paper against the stated inclusion and exclusion criteria: no control subjects (n=8); female or mixed-sex athlete data (n=30); non-English language (n=17); duplicate data sets (n=5); no absolute/useful cardiac data (n-14); drugs (n=30); mixed and unspecified sports (n=11); non-elite focus (n=7); review (n=1); inaccessible (n=3) and subjects outside age range (n=17). The remaining 101 studies were then scored by a single researcher (VU) using a quality criteria checklist specifically compiled for this systematic review (see online supplementary appendix 1). This was adapted from the STROBE,19 and PRISMA statements,20 to improve the systematic appraisal of the quality of observational studies. Nine records were excluded owing to low-quality scores, largely based on missing data related to training or athlete status and/or a lack of reference to professional guidelines for assessment and measurement. Consequently, 92 records were eligible for quantitative analysis, including 185 echocardiographic and 41 MRI datasets (see online supplementary appendix 2).

Data retrieval and meta-analysis

All relevant cardiac and body surface area (BSA) data were extracted directly from individual trials into a spreadsheet (Excel 2010, Microsoft Corp). Athlete groups and imaging modalities were coded discretely for each study. Continuous data for BSA as well as left ventricle, right ventricle and left atrium morphology and functional data were recorded as group mean±SD for each study. For the left ventricle we recorded left ventricular (LV) mass, interventricular septal wall thickness (IVSWT), posterior wall thickness (PWT), LV end-diastolic diameter (LVEDD), LV end-diastolic volume (LVEDV), LV stroke volume (LV SV), LV ejection fraction (LVEF), the ratio of LV peak early to atrial trans-mitral Doppler flow velocities (LV E/A), peak septal longitudinal tissue velocity in early diastole (LV E’), late diastole (LV A’) and systolic phase (LV S’). In the right ventricle we recorded right ventricular (RV) mass, RV end-diastolic diameter (RVEDD), RV end-diastolic volume (RVEDV) and RV SV. Finally, for the left atrium we recorded the left atrial dimension (LAD). Other parameters were considered at initial screening but too few papers recorded these data (eg, lateral LV wall tissue velocities).

To explore the impact of training group on structural and function parameters of the AH, we applied a mixed-effect, random meta-analysis model.21 To quantify study-to-study heterogeneity a Q statistic at p<0.05 and I2 statistic >50% was deemed significant.22 In further subgroup analyses we explored the impact of imaging modality upon AH data, again, using a mixed-effect, random meta-analysis model. Finally, we used a multiple meta-analysis regression model (Kendall's non-parametric statistic) to explore the impact of the covariate, BSA, on LV mass, RV mass and LAD. All statistical analysis was carried out with comprehensive meta-analysis software V.2.0 (Biostat, Englewood, New Jersey, USA) and Stata V.12 (Stata Corp, College Station, Texas, USA). Statistical significance was set at p≤0.05.


Across all studies the mean age of all male athletes and controls ranged from 18 to 38 years. The results of the impact of different training stimuli on various indices of LV structure and function are summarised in table 1. More datasets were available for endurance athletes and limited data prevented a mean pooled estimate for peak septal early diastolic tissue velocity in resistance athletes. All LV structural parameters were increased in athletes compared with sedentary controls. Differences between athlete groups were only noted for LVEDD and LVEDV, which were larger in endurance athletes than in resistance athletes. There were no differences between athlete groups for IVSWT or PWT. The larger LV chamber accounted for a greater SV in endurance athletes compared with resistance athletes and controls but LVEF did not differ among all groups. Both LV E/A and LV E’ were larger in endurance athletes than in controls.

Table 1

Left ventricular structural and functional data in male endurance-trained, resistance-trained and sedentary control subjects

Table 2 contains between-group comparisons for indices of RV structure and function as well as LAD. Noticeably, fewer studies have reported these data in cross-sectional athlete–control comparisons and we could not generate pooled mean estimates for resistance-trained athletes for RV data. Mean RV mass, RVEDV and RV SV were greater in endurance athletes than controls. Data for LAD were greater in endurance than resistance-trained athletes but not greater than in controls.

Table 2

Right ventricular structural and functional data as well as left atrial diameter in male endurance-trained, resistance-trained and sedentary control subjects

For many LV and RV variables there was significant evidence of study-to-study heterogeneity (see tables 1 and 2). This provided support for our a priori rationale to assess subsidiary factors. Substantial differences between imaging mode were noted (table 3). Pooled mean estimates for LV mass, LVEF and LAD were higher when using echocardiography. Conversely, pooled mean estimates for LVEDV and LV E/A were higher when using CMR. No imaging mode comparison was possible for the right ventricle because of limited data using echocardiography.

Table 3

Left ventricular structural and functional data as well as left atrial diameter in male athletes measured using echocardiography and cardiac magnetic resonance (CMR)


The key findings from this systematic review and meta-analysis of the AH in male athletes are that (1) both endurance and resistance-trained athletes demonstrate larger LV structures than sedentary controls with the greater dimensions in endurance athletes suggestive of an eccentric hypertrophy. Similar LV wall thickness in the two athlete groups provides minimal support for a concentric hypertrophy in resistance athletes; (2) the imaging mode has a significant, but inconsistent, effect on a range of LV indices and LAD; (3) differences between endurance athletes and controls were noted for LV function, RV structure and LAD, while limited resistance athlete data were available and (4) BSA has a significant positive relationship with LV mass, RV mass and LAD. These data should inform current knowledge of the AH and prompt continuing research.

Impact of training group

Both athlete groups had a larger LV wall, chamber dimensions and mass than the control group, which supports the existence of a morphological AH.5 The endurance-trained athletes had marginally larger LV mass and significantly greater LVEDD and LVEDV than resistance athletes, supporting the contention that endurance athletes tend to present with the largest LV dimensions.23 Furthermore, the pattern of LV morphology in the endurance-trained athletes, a bigger LV chamber and proportionately larger LV walls, is commensurate with an eccentric LV hypertrophy first proposed by Morganroth and coworkers.3 The mechanism(s) underpinning training-induced changes in LV morphology in endurance athletes are poorly understood but a haemodynamic volume overload is widely quoted.6

Endurance athletes had a larger LV SV than controls and resistance-trained athletes. This adaptation makes some teleological sense as an augmented LV SV is probably a key contributor to an enhanced endurance capacity.1 The lack of difference in LVEF at rest between all groups confirms data from Pluim et al,5 and suggests no between-group differences in contractility at rest. Both the LV E/A and LV E’ were significantly greater in endurance athletes than controls. Although controversial,5 individual studies have reported an improved diastolic filling at rest in athletes,24 yet this has often been dependent upon the specific parameter assessed.25 The potential importance of enhanced diastolic function in the development of maximal SV,26 as well as putative mechanisms (preload or intrinsic relaxation properties), requires further evaluation. Finally, it is interesting to note that the endurance-training-related changes in LV morphology and function are quite closely mirrored in RV and LA data. This supports a balanced cardiac hypertrophy that is assumed to be wholly physiological in nature.

Although there are noticeably fewer reported resistance-training studies, these data confirm the observation by Pluim et al5 that resistance athletes display some morphological characteristics of the AH. Both IVSWT and PWT were greater in resistance-trained athlete's controls and were similar to values in endurance athletes. Cavity dimension, but not volume, data were greater in resistance athletes than controls but were smaller than in endurance athletes. Pluim et al,5 noted some support for concentric hypertrophy in resistance athletes due to an increased wall to chamber ratio (relative wall thickness). Use of mean data from all groups in this meta-analysis results in a relative wall thickness of 0.40 for endurance athletes compared with 0.41 in resistance athletes. These data are within normal ranges,27 and not meaningfully different. As opposed to dichotomous cardiac structural responses to endurance and resistance training, it might be argued that both athlete groups present with a similar qualitative cardiac adaptation on a continuum, with greater cardiac dimensions in endurance athletes reflecting a greater overall training volume. The lack of concentric-type hypertrophy in resistance athletes might be due to (1) a limited exposure time to an increased haemodynamic afterload as an increase in blood pressure only occurs sporadically during resistance training because of the intermittent nature of repetitions, sets and work-to-rest ratios.23 The exposure to an increased haemodynamic load during exercise is probably much more consistent and substantial during endurance training; (2) the absence of any real afterload stimulus when resistance training is performed with a Valsalva manoeuvre.28 The resistance-trained athletes demonstrated no increase in resting LV SV or either index of diastolic function, compared with controls. Given that an enhanced SV is an unlikely contributor to resistance sports performance the lack of difference between the controls and athletes is not surprising. Data for RV and LA morphology and function are extremely limited in resistance-trained athletes and this requires further study.

Impact of imaging mode

There was strong evidence of study-to-study heterogeneity within the meta-analysis, suggesting that analysis of subfactors might be useful. CMR is the gold standard for morphological assessment of cardiac chambers and mass owing to its greater spatial resolution and three-dimensional data provision. Recent, direct comparisons between echocardiography and MRI-derived measures of LV mass and volume in athletes suggest that large absolute differences exist between these measurement modalities.29–31 Measurement variability is also substantially greater with echocardiography.11

In this study the use of CMR resulted in a higher LVEDV than echocardiography and this agrees with previous comparative studies.12 ,32–34 The difference is probably due to the biplane Simpson's technique that uses estimation and geometric modelling allied to poorer lateral resolution that makes clear delineation of the endocardium difficult. Conversely, LV mass, was greater when using echocardiography than with MRI, which supports previous work by Prakken et al,30 but contradicts other work.29 Because CMR provides a more accurate and precise measurement of LV mass,35 we can assume that echocardiography overestimates, again probably owing to the limitations of geometric assumptions,36 possibly compounded by the nature of any eccentric LV enlargement in endurance athletes.

These results, allied to the modality-related differences in LAD and LV function, highlight that imaging modalities should not be used interchangeably.37 For training studies or other within-subject testing at different times, a single modality should be used. Given the higher variability in echocardiographic estimates of LV dimensions, compared with CMR, some caution is warranted in the interpretation of small-sample echocardiographic AH studies. Further CMR studies in resistance athletes, or focusing on the right ventricle and left atrium, are required.

Impact of BSA

Finally, we sought to explore the effect of between-study differences in BSA on LV mass, RV mass and LAD. Multiple meta-regression data clearly showed that as BSA increased so did measures of cardiac structure. Consequently, it is likely that some portion of the between-group difference in LV mass, RV mass and LAD is due to the fact that those with larger cardiac dimensions have larger body dimensions. Whether this reflects a higher total body mass or, more specifically, a higher lean body mass in the athlete groups is impossible to determine from the use of BSA alone.

The importance of body size on cardiac dimensions has been demonstrated in a number of empirical studies,14 ,17 ,38 and has been highlighted in review articles.15 ,16 These data provide further support for the proposal that between-individual or between-group comparisons of cardiac dimensions must take into account individual variability in body size; otherwise, data interpretation and conclusions may be flawed. This study does not determine the best body size scaling index (mass, BSA, lean mass) or the most appropriate approach to scaling (ratio standards, allometry). These topics have been debated previously,15 ,16 and further empirical work is required. It is of critical importance that future AH studies report anthropometrics and/or scaled data.

Implications, limitations and future research

Strong evidence supports an ‘eccentric-type’ hypertrophy of the left ventricle, right ventricle and left atrium in endurance athletes. Resistance-trained athletes do not present with concentric LV hypertrophy. These data prompt a re-evaluation of the long-held belief that different exercise training produces divergent cardiac adaptation.3 A caveat is important here; we report substantial numbers of endurance-athlete groups but fewer resistance-athlete groups. This is an even greater problem for studies of RV and LA structure and function in athletes. Further work is required to confirm these suggestions in carefully selected groups of resistance athletes.

These findings also provide relevant information for those interested in the nature of the upper limits of human cardiac, physiological adaptation to training. It is likely that the upper limits for chamber dimensions will be seen in endurance athletes. This knowledge will inform cardiac screening and the differential diagnosis of AH from pathological adaptation. Further, it is unlikely that the upper limits of LV wall dimensions will be seen exclusively in resistance athletes. It is also important to emphasise that absolute wall thicknesses and the LV end diastolic dimension, although increased in comparison with sedentary controls, do not fall within the pathological range seen in hypertrophic or dilated cardiomyopathy in either resistance- or endurance-trained athletes. This knowledge will further aid the diagnostic challenges associated with preparticipation screening of the competitive athlete.

Continuing study in the field of the AH should, where possible, use CMR to determine cardiac morphology. Where CMR is not available echocardiography studies should be adequately powered. The assessment of LV, RV and LA function requires substantial development. Here CMR is ably supported by newer echocardiographic technologies such as tissue-Doppler and speckle tracking.39

In this meta-analysis, we examined the influence of BSA on sample differences in cardiac dimensions via a multiple meta-regression approach. Although this analysis goes some way towards partitioning the influence of BSA on the cardiac dimension estimates, proper allometric scaling approaches should, ideally, be applied to individual studies. Unfortunately, this approach to analysis is rare.

This study has some limitations. As already noted, data for resistance athletes are scarce. Functional measures in high-quality case–control series studies are largely limited to global echocardiographic parameters—namely, LVEF, SV and E/A. This study excluded older athletes and all female athlete studies. These limitations prompt continuing evaluation in this area.


This large-scale systematic review and meta-analysis in male athletes provides strong evidence of LV, RV and LA hypertrophy with athletic training that is not dichotomous, but is quantitatively greater with endurance training. Significant evidence of study-to-study heterogeneity was noted that might be due to the use of different imaging modalities and the approach to scaling (or indexing) cardiac structural data for individual differences in body size. Consequently, this meta-analysis provides a useful re-evaluation of concepts and models in the AH literature.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors KG, DO and JS conceptualised the idea. All authors contributed to the research design, inclusion/exclusion criteria. VU and KG carried out the database searching and initial screening. VU compiled the final data tables and quality criteria scoring. GA and VU completed the data analysis. All authors contributed to the reporting of the work. VU and KG are responsible for the overall content as guarantors.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.