Background
An increasing number of people ascend to altitude each year for recreational, occupational, and sporting purposes [
1]. The hypoxic exposure experienced at altitude is known to cause a curvilinear impairment in endurance performance with increasing levels of hypoxia [
2]. However, the changes in substrate utilisation associated with these decrements in performance are currently unclear, with some authors reporting an increased contribution of carbohydrate to the total energy yield [
3,
4], and others demonstrating an increased contribution of fat oxidation [
5,
6]. Developing a better understanding of these changes in substrate utilisation in hypoxia is vital in designing dietary interventions to maintain and/or improve performance in such conditions.
Exposure to hypoxic environments may alter substrate utilisation through multifarious mechanisms. It is purported that the mechanism responsible for increased carbohydrate dependency in hypoxia is mediated by the sympathetic nervous system, via the secretion of epinephrine and norepinephrine, stimulating glycogenolysis and gluconeogenesis [
3,
7]. Alternative evidence suggests that increases in the transcription factor hypoxic-inducible factor 1 alpha (HIF-1α) may be responsible for the increases in fat oxidation observed by some in hypoxia, via upregulation of the fatty acid-activated transcription factor peroxisome proliferator-activated receptor alpha (PPARα) [
8]. Albeit in rats, PPARα has been demonstrated to deactivate pyruvate dehydrogenase [
9], inhibiting the conversion of pyruvate to acetyl-coA and therefore enabling greater fat flux for oxidation [
10].
A myriad of factors has been suggested to influence the interaction between hypoxic exposure and substrate utilisation. These include, but are not limited to: characteristics of hypoxic exposure (severity, duration and type) [
3], nutritional status of participants (fasted or fed before exercise/exogenous supplementation during exercise) [
4,
5] and sex of participants [
11]. In relation to hypoxic exposure, hypobaric hypoxia (HH) has been suggested to elicit more severe physiological responses (greater hypoxemia and lower arterial oxygen saturation) compared with normobaric hypoxia (NH) [
12]. Although this is contested in the literature [
13,
14] it seems plausible that these physiological differences may induce a greater reliance on carbohydrate oxidation to achieve a higher yield of ATP per unit of oxygen consumption, compared with fat oxidation [
15]. This theory may also be applied to the effect of altitude severity on subsequent substrate oxidation. Further, metabolic responses may be different between sex, with females demonstrating a greater relative utilisation of fat oxidation [
11,
16]. This may be attributable to a number of factors, with females demonstrating a greater relative fat mass [
17] and intramuscular triglyceride stores [
18], as well as better mobilisation of free fatty acid (FFA) from subcutaneous adipose tissue [
19]. It has also been suggested that this propensity for fat oxidation may be mediated by the steroid hormones estrogen (predominantly 17 β-estradiol) and progesterone [
20,
21]. Regarding nutritional status of participants, equivocal metabolic findings have been observed in response to carbohydrate supplementation during exercise in hypoxia, compared with normoxia [
4,
5], which demonstrates the limited current understanding of the interaction between dietary interventions and hypoxic exposure.
To gain a clear understanding of changes in substrate utilisation during exercise in hypoxia compared with normoxia, a systematic evaluation is required to explain the equivocal results of previous studies. Due to the greater exercise-induced physiological stress experienced when performing a matched absolute workload under hypoxic conditions [
22], this meta-analysis focuses solely on exercise matched to relative intensities. The aim of this meta-analysis was to identify the study characteristics responsible for heterogeneity between findings, using subgroup analyses and meta-regression.
Methods
The current systematic review and meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic Review and Meta-analyses (PRISMA) guidelines [
23].
Literature search
A literature search was conducted using the electronic bibliographic databases PubMed and the Cochrane Library, as well as searching MEDLINE, SPORTDiscus, PsychINFO, and CINAHL via EBSCO
host. The initial search of titles, abstracts and keywords was conducted on 15th November 2016 using terms related to ‘exercise’, ‘hypoxia’, ‘substrate’ and ‘oxidation’. A final search was conducted on 5th June 2018. The specific keywords and full search strategy can be found in Additional file
1. The reference lists of all included studies and relevant review articles were screened for possible inclusion. No language restrictions were applied and in the case of studies available only as an abstract, authors were contacted for the full dataset.
Inclusion criteria
Included studies were required to meet the following criteria: participants in the study were between the ages of 18 and 65 years, not pregnant, non-smokers, with no history of diabetes, gastrointestinal, inflammatory, metabolic, cardiovascular, neurological or psychological disease. In order to minimise potential publication bias, studies published in peer reviewed journals, conference proceedings, theses or dissertations were eligible for inclusion.
All studies were required to measure RER and/or carbohydrate or fat oxidation. These measures were required to be quantified during exercise matched for relative intensities in hypoxic and normoxic environments. Hypoxic exposure was defined as terrestrial altitude via geographical location (TA) or simulated altitude (NH or HH) via a hypoxic tent, hypoxic chamber or breathing mask. Exposures were required to be > 1500 m or a simulated equivalent (i.e., low altitude or higher) [
24]. All participants within selected studies had not been exposed to > 1500 m (or a simulated equivalent) within the previous 3 months. Normoxic trials were required to provide a viable within-participant control (i.e. equivalent measure(s) quantified in the same participants as a separate trial in normoxic conditions). The exercise was required to be > 5 min in duration to achieve steady-state values at a fixed exercise intensity [
25].
Two researchers (AG and OS) independently assessed studies for inclusion and later compared notes to reach a mutual consensus. Disagreements about the eligibility of any particular studies were resolved by a third reviewer (KD). Potential studies that could not be excluded based on their title or abstract were retrieved in full-text and reviewed against the inclusion/exclusion criteria independently by two researchers (AG and OS) with a third researcher (KD) used to settle any disputes. In total, 18 studies met the inclusion criteria and were included in this meta-analysis.
Abstraction of data
Data were extracted independently by two researchers (AG and OS) into a standardised spreadsheet, which included (i) characteristics of articles valid for review; (ii) the Cochrane Collaboration’s tool for assessing risk of bias, and (iii) outcome data suitable for analysis based on mean, standard deviation (SD) and sample size. Further data was extracted regarding participant characteristics, acclimatisation status, nutritional manipulations, exercise intensities and duration, exercise mode, and severity and duration of hypoxic exposure. In studies which employed multiple exercise intensities, each respective intensity was directly compared with the equivalent intensity in the alternate condition.
In studies which reported outcome variables across numerous time points during exercise, values were averaged to calculate the mean. In addition, SD values were averaged using the following formula:
$$ \frac{{\mathrm{n}}^1\left(\mathrm{S}{1}^2+\mathrm{D}{1}^2\right)+{\mathrm{n}}^2\left(\mathrm{S}{2}^2+\mathrm{D}{2}^2\right)\dots }{\left({\mathrm{n}}^1+{\mathrm{n}}^2\right)\dots } $$
where:
n1 = sample size of group 1
n2 = sample size of group 2
S1 = SD of group 1
S2 = SD of group 2
D1 = mean of group 1 – mean of total group
D2 = mean of group 2 – mean of total group
… denotes inclusion of further data points if required
Absolute substrate oxidation data was converted to g·min− 1. Thus, values expressed as total grams oxidised throughout exercise were divided by the number of minutes the variable was measured. Values expressed in mg·kcal·min− 1 were multiplied by 1000 to convert to g·kcal·min− 1, and then multiplied by kcal values provided in the relevant paper for the conversion to g·min− 1. In addition, carbohydrate oxidation data provided in mmol·min− 1 were divided by 1000 and then multiplied by the molar mass of glucose (180.1559 g/mol). Where values were presented as figures, these were digitized using graph digitizer software (DigitizeIt, Germany) and the means and SD were measured manually at the pixel level to the scale provided on the figure.
Assessment of risk of bias in included studies
Two independent reviewers (AG and OS) used The Cochrane Collaboration’s tool for assessing risk of bias [
26] to determine the risk of bias in each study. Each study was assessed in the following six domains: sequence generation, allocation concealment, blinding of participants, personnel and outcome assessors, incomplete outcome data, selective outcome reporting and other sources of bias (e.g. has been claimed to have been fraudulent). A judgement was made on each of the domains by the two independent researchers as to whether they were ‘high risk ‘or ‘low risk’. When insufficient detail was reported then the judgement of ‘unclear risk’ was made. Disagreements were solved initially via discussion between the two independent reviewers however a third reviewer (KD) was consulted for dispute resolution. ‘Risk of bias graphs’ were computed in Review Manager (RevMan) 5.3 (The Cochrane Collaboration) to include low, unclear and high risk for each domain.
Statistical analysis
Outcome measures were quantified using mean difference between conditions with 95% confidence intervals (CI) which were used as the summary statistic. A random-effects meta-analysis was performed by AG, JM and KD using Comprehensive Meta-Analysis Software (version 3, Biostat, Englewood, NJ, USA). The inputted data included sample sizes, outcome measures with their respective SDs, and a correlation coefficient for within-participant measurements. These correlation coefficients were estimated from prior studies in our laboratory and were as follows: RER r = 0.78, absolute carbohydrate oxidation r = 0.70, absolute fat oxidation r = 0.81, relative carbohydrate oxidation r = 0.79, relative fat oxidation r = 0.79 [
5,
6].
A negative mean difference indicates that hypoxic exposure was associated with a decrease in the respective outcome variable, while a positive mean difference indicates that hypoxic exposure was associated with an increase in the respective outcome variable. Heterogeneity between trials was assessed using the Chi-squared statistic, I-squared statistic and the Tau-squared statistic.
To examine whether any conclusions were dependent on a single study, sensitivity analyses was employed for each variable by repeating the analyses with each study omitted in turn.
Where the number of comparisons was suitable, meta-regression analysis was performed. This analysis was used to determine whether continuous or categorical data, including severity of hypoxic exposure, exercise intensity and pre-exercise nutritional state could explain the variation in the values observed between studies. Each moderator was analysed in a meta-regression independently to determine their relationship with the outcome variable. Where significant moderators were identified, multiple meta-regression analysis was conducted to determine if these remained significant, whilst controlling for all other moderators. All meta-regressions were performed using the restricted maximum likelihood (REML) method with Knapp-Hartung adjustment.
Duration of hypoxic exposure was categorised into acute and chronic as per each study’s description of their own exposure (acute < 44 h; chronic = 3–28 days). Overall mean differences, CIs and p values were reported for all variables during exercise matched to relative intensities. Meta-regression analysis was performed only on the outcome variable RER, as this represents the most appropriate measure to assess changes in the relative substrate contributions of carbohydrate and fat (i.e., physiological shifts in substrate utilisation). Meta-regression analysis of absolute fat and carbohydrate oxidation rates was deemed unnecessary based on these responses determining the RER values for each study.
Exploration of small study effects
Small study effects were explored using funnel plots of mean difference versus standard errors [
26], and by quantifying Egger’s linear regression intercept. A statistically significant Egger’s statistic (
p < 0.05) indicates the presence of small study effects.
Discussion
The purpose of this systematic review and meta-analysis was to examine the effects of hypoxic exposure on substrate oxidation during exercise matched to relative intensities. There was no consistent change in relative carbohydrate or fat contribution to energy provision during exercise matched for relative intensities in hypoxia, compared with normoxia. These findings are particularly pertinent as, in contrast to exercise matched to absolute intensities, exercise matched to relative intensities isolates the effect of hypoxia by normalising the exercise intensities between conditions [
22]. Additionally, the heterogeneous findings of the current literature may be explained by a number of differing experimental characteristics, such as pre-exercise nutritional status and exercise intensity.
There was no significant change observed in RER during exercise matched for relative intensities in hypoxia, compared with normoxia. In addition, there was no significant change in relative carbohydrate or fat oxidation in the same circumstances. It was deemed that RER was the most useful outcome variable due the largest number of comparisons as a result of being the most frequently reported in the literature. As expected, reduced absolute carbohydrate rates were observed in hypoxia during exercise matched to relative intensities, due to the lower absolute workload [
27] performed in hypoxia than normoxia and the subsequent reduction in energy expenditure. However, no significant change in absolute fat oxidation was observed in hypoxia, likely due to the limited changes in the contribution from this fuel source at moderate (40–55% VO
2max) exercise intensities [
28].
The current review found that an increase in RER was induced during exercise matched for relative intensities in hypoxia compared with normoxia when participants were in the fed state. Alternatively, a decrease in RER was induced when in the fasted state. The increase in endogenous carbohydrate stores as a result of feeding may facilitate a hypoxic-induced physiological demand for increased carbohydrate oxidation, thus potentiating the fuel shift. The mobilisation and oxidation of these stores may be augmented by the synergistic effect of feeding [
29] and hypoxia [
3] on sympathetic nervous system activity (i.e. increased secretion of epinephrine and norepinephrine) and resultant increases in gluconeogenesis and glycogenolysis. Interestingly, recent evidence also suggests that the rise in circulating insulin concentrations after feeding may increase carbohydrate oxidation from muscle glycogen stores, even before the ingested carbohydrate has been transported into the muscle [
30]. This is supported by previous work demonstrating a reduction in muscle glycogen concentrations one-hour after consuming a mixed macronutrient meal, before increasing again in the subsequent hours [
31]. This increase in insulin concentrations after a pre-exercise meal may be potentiated by hypoxia [
6], thereby enhancing the inhibition of lipolysis and FFA mobilisation [
32] to increase carbohydrate oxidation. In contrast, fasted exercise may elicit a decrease in RER via the enhanced activation of PPARα due to both hypoxia [
8] and fasting [
33]. The synergistic effect of both factors may further disrupt glycolysis [
9] and enable greater fat flux [
10].
In addition, an increased RER was observed during exercise matched to relative intensities in hypoxia, compared with normoxia, during exercise performed at higher intensities. This effect may be mediated by the hypoxic effect of altitude and high intensity exercise, augmenting skeletal muscle hypoxia [
34]. The mechanisms associated with these changes are likely explained as per the physiological response to increased exercise intensities in normoxic environments. In this regard, higher exercise intensities induce a reduction in adipose tissue blood flow, which may attenuate the release of FFA resulting in decreased delivery to the contracting muscle [
35]. Further, greater exercise intensities stimulate greater flux through the glycolytic pathway and pyruvate dehydrogenase complex (PDC) than flux through the tricarboxylic acid cycle, resulting in the accumulation of acetyl coA [
36]. The subsequent acylation of the carnitine pool has been suggested to result in a marked decrease in muscle free carnitine and downregulation of carnitine palmitoyltransferase I (CPT-1), the enzyme responsible for transporting long chain fatty acids into the mitochondrial matrix [
28]. Alternatively, the effect of hypoxia on the sympathetic nervous system may be potentiated by greater exercise intensities, enabling greater carbohydrate oxidation due to increased glycogenolysis, a result of enhanced glycogen phosphorylase activity, sarcoplasmic Ca
2+, inorganic phospohate and cyclic AMP [
37,
38]. Numerous mechanisms are proposed to explain the reduction in FFA oxidation with increasing exercise intensities [
39], however detailed discussion of all theories is out of the scope of this review.
The large between study heterogeneity in relation to RER during exercise matched to relative intensities was explained in part by pre-exercise nutritional state and exercise intensity (~ 42%). The remaining, unexplained heterogeneity may highlight some limitations of the present meta-analysis. Results from a meta-regression are indicative of a between-study relationship, however due to confounding bias (i.e. one experimental characteristic may reflect a true association with other correlated, known or unknown characteristics), this relationship may not be replicated within-studies. This is termed aggregation bias. As such, moderator analysis should be regarded as hypothesis gathering, rather than hypothesis testing [
40]. These moderators should therefore subsequently be investigated using a within-measures design via randomised controlled trials, generating causal, rather than observational relationships. Further, the unexplained heterogeneity may be due to methodological heterogeneity (i.e. study quality/measurement error) or insufficient trials to generate the appropriate power to fully explain the heterogeneity. A greater quantity and quality of research regarding substrate oxidation during hypoxia would help to further explain the heterogeneity between trials Further research is required to confirm the findings from this meta-analysis and quantify the influence of the fasted and fed state and exercise intensity on substrate utilisation in hypoxia.
The present meta-analysis provides clarity, and therefore facilitates an accurate interpretation, of the current literature. These findings may inform nutritional strategies for mountaineers, military personnel and athletes during exposure to altitude, subsequently limiting the detrimental exercise performance experienced in such conditions. The performance benefits of maintaining exogenous carbohydrate oxidation and/or endogenous carbohydrate stores via pre-exercise carbohydrate consumption in normoxia are well documented [
41]. As such, findings from this review suggest that a physiological drive for carbohydrate oxidation in hypoxia may be facilitated by an increased carbohydrate intake prior to exercise, in order to avoid an accelerated depletion of muscle glycogen, and shift back to the less efficient oxidation of fat [
35]. In contrast, the use of low carbohydrate intake strategies to enhance endurance training metabolic adaptations is growing in popularity [
42] and findings from the present meta-analysis may have implications for such strategies. Specifically, the combined effect of training in hypoxia in a glycogen depleted state may potentiate the metabolic adaptations of ‘training low’. Alternatively, a number of studies have demonstrated that changes in substrate utilisation during exercise in hypoxia may have implications for metabolic disease programmes [
43,
44]. As such, the clinical translation of the current study warrants further investigation.
Despite the important findings observed in the current meta-analysis, some notable limitations must be acknowledged. First, the equivocal findings observed in RER and relative substrate oxidation reflects the heterogeneity in the literature, rather than an absence of change in substrate oxidation due to hypoxic exposure. This heterogeneity is likely due to differing experimental characteristics between studies and although moderator analysis was employed to identify these factors, these findings should be interpreted with caution. Second, the physiological determinants of substrate oxidation (e.g. hormonal factors), were not quantified and therefore physiological mechanisms were difficult to elucidate. This was, however, beyond the scope of this study. Finally, despite an extensive search returning 1743 records, we cannot guarantee that our search was completely exhaustive of the relevant literature. However, should the primary or secondary aim of a study be related to this area, they are likely to have been detected in our search.