FormalPara Key Points

The energetic cost of cycling may be improved by exercise training in healthy humans.

Exercise programs including strength training improve the energetic cost of cycling in previously trained or untrained subjects.

Endurance training is only effective at improving the energetic cost of cycling in previously untrained subjects.

1 Introduction

Humans exhibit a gain in oxygen uptake (VO2) of 14–25 mL·min−1·W−1 during submaximal cycle ergometer exercise [1–18], which roughly corresponds to 12–21 % mechanical efficiency as determined by the ratio of external work to total energy expenditure [19]. In the context of any activity limited by the capacity to expend energy, such as exercise performance, a high exercise economy and efficiency, as represented by low VO2and caloric cost during submaximal exercise, respectively, is of paramount importance [20–23]. Hence, there is a particular interest to identify potential determinants of and strategies to reduce the energetic cost of exercise (EE) [24–26].

Whole-body EE is determined by the combination of efficiencies along the chain of energy transformation and relates to biochemical, physiological, anatomical, and biomechanical factors [27, 28]. Given the relatively fixed body position during cycling, biomechanical aspects are seemingly a less relevant component of the energetic cost of cycling (CE) than running (RE), which might explain, in part, the lower between-individual variability in CE vs. RE [20, 26]. In contrast to other major determinants of exercise performance (e.g., maximal oxygen consumption, lactate threshold), relatively little is known regarding the trainability of EE [29]. It seems reasonable to expect changes in EE following interventions inducing widespread adaptations such as exercise training; indeed, a range of exercise interventions are suggested to improve RE [30]. However, it remains unclear whether CE is improved with exercise training [20, 31, 32], despite related research spans over 4 decades [1–18, 22, 33–48]. Presumably, the small sample size as well as distinct training characteristics, study population, and methodology of individual studies may have compounded the impact of exercise training on CE [1–18, 22, 33–48]. In this regard, a meta-analytical approach may contribute to clarify the effect of exercise training on CE, but to our knowledge, this has not yet been performed.

Therefore, the purpose of this study was to perform a systematic review and meta-analysis on the effect of exercise training on CE in healthy subjects as well as to identify potential moderating/confounding factors.

2 Methods

The review was conducted according to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) Group guidelines [49].

2.1 Data Sources and Searches

Our systematic search included the databases MEDLINE, Scopus, and Web of Science, from inception until December 2014. We used combinations of the subject headings ‘healthy’, ‘training’, ‘exercise’, ‘efficiency’, and ‘economy’; the search strategy for MEDLINE is shown in Fig. S1. We also performed hand searching in reviews identified through the systematic search, articles included in the meta-analysis, related citations in MEDLINE, personal bibliography, and Google.

2.2 Article Selection

To be included in the analysis, an original research article had to assess CE before and after an exercise training intervention in healthy subjects. Studies were excluded if they complied with the above criteria but involved other interventions deemed likely to influence CE. Likewise, studies were excluded if they assessed CE only at exercise intensities above the LT, to limit the potential confounding influence of anaerobic energy systems on the effect of exercise training on CE. In addition, if VO2 values during the CE test were divided by weight, the latter had to be not significantly altered by the training intervention. In the event of multiple publications pertaining to the same research, the first published or more comprehensive report was included. Inclusion of articles in our analysis was not limited by publication status or language.

2.3 Data Extraction and Quality Assessment

The following variables were summarized in a pre-formatted spreadsheet: authors, year of publication, characteristics of study participants (n, age, sex, height, weight, body fat, lean mass, body mass index, ventilation, heart rate, stroke volume, cardiac output, arteriovenous oxygen difference, lactate, blood pressure, vascular peripheral resistance, blood volume, red cell volume, Hb concentration, hematocrit, fiber-type distribution, mitochondrial content, muscle capillarization, maximal oxygen consumption, respiratory exchange ratio, maximal power with incremental exercise, maximal voluntary contraction, ratio of perceived exertion, fitness status, nutritional status, health status), exercise training features (type, modality, intensity, session length, frequency, duration), and characteristics of the CE assessment (workload, length, cadence, units of measurement). In the case of concurrent reports of CE at different workloads, the CE assessment at the lowest workload, after 5 min of warm-up, was used in the meta-analysis [6, 11, 14, 18, 22, 41, 43]. In the case of concurrent VO2 (l·min−1) and other units of measurement of CE, the former was used in the meta-analysis given its prevailing report, to attenuate the methodological variability between studies [1, 4, 44]. A systematic appraisal of quality for observational research (SAQOR) [50], previously applied in meta-analysis of observational studies evaluating the effect of exercise training [51], was performed to determine study quality. The SAQOR was adapted to assess (1) the study sample, (2) quality of CE assessment, (3) confounding variables, and (4) data. Overall, the SAQOR was scored out of 15, quality deemed better with a greater score (0–5 low, 6–10 moderate, 11–15 high).

2.4 Data Synthesis and Analysis

The meta-analysis and related analyses were performed using Review Manager software (RevMan 5.3, Cochrane Collaboration, Oxford, UK) and Comprehensive Meta-analysis software (version 2, Biostat, Englewood, NJ, USA). The primary outcome was the standardized mean difference (SMD) between post- and pre-training measurements in CE. The SMD summary statistic allowed us to standardize values obtained using different methods into a uniform scale to complete the meta-analysis [52]. Each SMD was weighted by the inverse variance and they were pooled with a random-effects model [52, 53]. According to Cohen’s conventional criteria [54], SMD of 0.2, 0.5, and 0.8 represents small, medium, and large effect sizes, respectively.

Heterogeneity among studies, defined as the variation in the intervention effects that are not compatible with chance alone, was assessed using the chi-squared test for heterogeneity and I 2 statistics. Potential moderating/confounding factors influencing the SMD in CE were evaluated by subgroup analysis comparing studies grouped by dichotomous variables. In addition, meta-regression analyses were performed to evaluate the associations between the SMD in CE and quantitative variables. In all meta-regression models, studies were weighted by the inverse variance of the dependent variable. Potential moderating/confounding factors were entered as independent variables in regressions models with the SMD in CE as the dependent variable. A negative association represents an increased positive effect of training on CE correlated with higher values of the associated variable, and vice versa. Publication and/or other biases were evaluated by the Begg and Mazumdar’s rank correlation test, Egger’s regression test, and visual inspection of funnel plot symmetry [52, 55]. A P value of less than 0.05 was considered statistically significant.

3 Results

3.1 Study Selection and Characteristics

The process of article selection is illustrated in the flow diagram (Fig. 1), which resulted in the inclusion of 35 articles. Twelve of these articles comprised separate study groups [1, 2, 7, 11, 14–16, 34, 35, 38, 39, 48], each of which was evaluated as an individual study in the meta-analysis. Table 1 illustrates the main characteristics of the included 51 studies, comprising a total of 531 healthy (mostly male) subjects with a mean age ranging from 20 to 66 years. Twenty-six studies involved previously untrained subjects, 17 studies comprised trained subjects, while 8 studies did not report fitness status. Among the latter, none reported high VO2max levels according to age- and sex-adjusted guidelines [56]. The majority of the training interventions consisted of endurance- and/or strength-structured training programsof variable intensity performed by means of cycle ergometry, treadmill/running, and/or resistance exercise, ranging from 0.71 to 17 h per week and from 4 to 34 weeks of duration (Table 2). The quality of the studies was moderate. The mean score was 7.4 ± 2.1 out of a possible 15 points (Table S1). As for the evaluation of potential biases, the funnel plot (Fig. S2), Begg and Mazumdar’s rank correlation test (P = 0.62), and Egger’s regression test (P = 0.81) suggested the absence of publication and/or other biases for the SMD in CE in the studies included in the meta-analysis.

Fig. 1
figure 1

Flow diagram of the process of article selection. CE cycling energetic cost

Full size image
Table 1 Main characteristics of studies included in the meta-analysis
Full size table
Table 2 Exercise training characteristics of studies included in the meta-analysis
Full size table

3.2 Effect of Exercise Training on Cycling Economy (CE)

CE was determined in all studies during cycle ergometer exercise of low, moderate, or moderate-to-high intensity (Table 1). CE was predominantly expressed as VO2 (l·min−1) at a given submaximal workload (i.e., ‘gross’ CE). After data pooling, the meta-analysis revealed an improved CE after training (SMD = −0.28, 95 % confidence interval = −0.44, −0.13; P = 0.0004) (Fig. 2). Heterogeneity was detected between studies (I 2 = 36 %, P = 0.007). In subgroup analyses (Table 3), there was no effect of training on CE in studies including only female subjects (n = 4) or previously trained subjects (n = 17) (P = 0.10 and P = 0.19, respectively). Likewise, studies applying only endurance training (n = 33), assessing CE with free/unknown cadence (i.e., pedaling rate) (n = 16), or through multiple workloads (n = 11) did not result in improved CE after training (P = 0.08, P = 0.10, and P = 0.25, respectively). In previously trained subjects, CE was improved only in studies applying strength training (n = 11), while in previously untrained subjects, both endurance (n = 20) and strength training alone or along with endurance training (n = 5) resulted in improved CE. Nonetheless, none of the aforementioned subgroup analyses, except for the comparison between types of training in all subjects, revealed a significant difference between subgroups (P Difference > 0.05).

Fig. 2
figure 2

Forest plots of the standardized mean difference (SMD) between post- and pre-training measurements in CE. Squares the SMD for each study. Diamonds represents the pooled SMD across studies. CI confidence interval, CE cycling energetic cost, df degrees of freedom, IV inverse variance, SD standard deviation

Full size image
Table 3 Subgroup analyses of the effect of exercise training on CE
Full size table

3.3 Meta-Regression Analyses

Table 4 presents the associations between the SMD in CE and potential moderating/confounding factors. Considering all studies included in the meta-analysis, the SMD in CE was (1) positively associated with ∆body fat (B = 0.31, 95 % CI [0.11, 0.50]) and cadence (B = 0.04, [0.02, 0.06]) in the CE test, and (2) negatively associated with total hours (B = −0.00, 95 % CI [−0.01, −0.00]) and duration (B = −0.03, [−0.05, −0.01]) of training and pre training VO2 (B = −0.53, [−0.95, −0.11]) in the CE test. These and further associations were observed when potential moderating/confounding factors of the SMD in CE were assessed separately according to the training status of study participants or the type of training intervention implemented (see details in Table 4).

Table 4 Significant associations between the standardized mean difference (SMD) in CE and potential moderating/confounding factors, as determined by meta-regression
Full size table

4 Discussion

In this systematic review and meta-analysis, we pooled and analyzed data from 51 studies assessing the effect of exercise training interventions on CE in a total of 531 healthy humans across a wide range of ages. The key observations of this analysis are: (1) strength training alone or along with endurance training improves CE and shows (2) a superior effect compared with endurance training alone; in turn, (3) endurance training alone is only effective at improving CE in previously untrained subjects.

Among the major contributors to endurance cycling performance, CE is considered the slowest responsive factor, if at all, to training [20, 57]. Nonetheless, given that even minor improvements in CE may yield substantial increases in performance [58], interventions/training regimes that may enhance CE are of great interest [29]. Herein, the meta-analytical evidence demonstrates that exercise programs including strength training improves CE to a moderate degree in previously trained and untrained healthy subjects (Fig. 2; Table 3). This agrees with the established improvement in RE following strength training [24]. Provided a negligible influence of biomechanics on CE, it seems unlikely that the effect of strength training on CE are related to biomechanical factors. Putative mechanisms by which strength training may enhance CE fundamentally comprise improved neuromuscular coordination [59, 60] and increased force per motor unit leading to reduced muscle activation at a given workload [61]. Importantly, none of the studies included in this meta-analysis reported the inclusion of strength-trained subjects (Table 1), in whom strength adaptations and possibly CE improvements might be of lesser magnitude [62]. Moreover, there might be quantitatively and qualitatively distinct adaptations and thereby CE modification in accordance with different modalities of strength training [63]; however, limited data availability precluded us to separately analyze the impact of strength training modalities in the present work. Regardless, the low variability of the effect of strength training alone or along with endurance training on CE is noteworthy (I 2 = 13 %, P = 0.31) (Fig. 2), suggesting similar CE responsiveness to exercise interventions including strength training.

On the basis of cross-sectional studies, endurance training is believed to have no influence on CE [57, 64, 65]. Yet, case reports in elite athletes have shown marked improvements in CE with years of endurance training [66, 67]. The present study reveals a small effect of endurance training (lasting 1–4 months) on CE in previously untrained, but not trained, subjects (Table 3). This suggests that (1) the nature and/or extent of typical short-term adaptations to endurance training such as increased oxygen delivery [68], muscle capillarization and mitochondrial content [69], among others, have little impact on CE and (2) longer endurance training interventions may be required to improve CE in physically fit humans. With respect to the latter, it has been hypothesized that the enhancement of CE with long-term endurance training is attributed to large increases in the percentage of type I muscle fibers [67], although related experimental evidence in humans is lacking. Additionally, changes in blood flow distribution following years of endurance training might influence CE [70]. In this regard, endurance athletes exhibit increased oxygen extraction, lower limb blood flow, and similar vascular conductanceper submaximal workloadcompared with untrained peers [70–72], conceivably due, in part, to a more efficient limb blood flow distribution towards exercising skeletal muscle [70]. This may contribute to lower myocardial work and improve CE, albeit onlyto a minor degree, in endurance athletes. Otherwise, long-term metabolic adaptations to endurance training [73–75] do not seem to affect the VO2/ATP production ratio component of CE because mithocondrial efficiency is similar over a wide range of fitness conditions [21, 25, 74]. In turn, the endurance training-induced improvement in ‘metabolic stability’ (i.e., reduced changes in the concentrations of muscle metabolites such as ADP, AMP, inosine monophosphate, creatine, inorganic phosphate, and H+ for a given ATP turnover) may be crucial to limit muscle fatigue, VO2 slow component, and CE impairment occurring at heavy and severe exercise intensities, particularly through a decrease in the ATP use/power output ratio [21–23].

The observation that CE was enhanced with the duration of the exercise intervention might challenge the independent status of the above findings. Yet, in the included studies no difference was detected for the weighted average duration of training according to type of training (P = 0.12) and fitness status (P = 0.18). It should be noted that there was no association between the SMD in CE and the duration of the intervention in studies including trained subjects, whereas a linear association was found in studies including untrained subjects (Table 4). This may denote the presence of a duration threshold for the impact of training on CE in previously trained subjects. In addition, several anthropometrical and methodological characteristics were associated with the effect of training on CE when considering all studies (Table 4). For instance, the higher the submaximal VO2 during baseline CE assessment, the greater the favorable effect of training was on CE. This could suggest an increased susceptibility for CE improvement in subjects with low CE at baseline. Likewise, the decrease in body fat percentage with training was directly related to the improvement in CE. Whilst speculative, a lower body fat percentage, if accompanied by reduced leg mass, may reduce CE through a decreased cost of leg movement. Furthermore, we detected an enhanced positive effect of training on CE with lower pedaling rates during the CE assessment. This adds to previous meta-analytical findings showing a worsened CE with higher cadence in un- to highly trained subjects [32], which also may imply that potential changes in CE may have been overlooked if determined with higher vs. lower cadences. Ultimately, given the nature and limited bivariate design of the heterogeneity analyses here applied, all associations should be considered exploratory and not as proof of causality [76].

There are additional limitations in this systematic review and meta-analysis worth addressing.

First, the majority of the included studies (44 out of 51) used VO2 during submaximal cycling at a given workload, i.e., economy, as a measure of CE (Table 1), thus the latter could have been modulated by training-induced adaptations in substrate metabolism (e.g., increased fat use and higherVO2/ATP production ratio). Nonetheless, there was no correlation between the SMD in CE and training-related changes in respiratory exchange ratio during the CE assessment (19 studies). This suggests that potential adaptations in substrate use did not affect the results of this meta-analysis, as indeed was also supported by the similar SMD in CE between studies that expressed CE in economy vs. efficiency units (P = 0.50). Second, studies assessing CE only at exercise intensities reasonably above the LT were excluded. This could have underestimated the effect of exercise training on CE because greater attenuation of the VO2/W slope above vs. below the LT after training has been reported [22, 41, 43]. Third, few studies comprised female individuals, which hinders any conclusion on the effect of sex on CE. Finally, the methodological quality of the included studies was determined as moderate, although there was no evidence of publication bias (Fig. S2).

5 Conclusions

The current meta-analysis demonstrates that exercise training may improve CE in healthy subjects. Such improvement is observed with exercise programs including, albeit not restricted to, strength training in previously trained subjects. In untrained subjects, CE is also enhanced by endurance training. These data therefore buttress the addition of strength training to any exercise program aiming to enhance CE and thereby endurance performance. Further research is needed to establish the most effective exercise strategy to improve CE in athletic populations.