Design
The Consideration of Population, Intervention, Comparator, Outcomes and Study design (PICOS) framework was used:
Population
Included participants had to fit the following criteria: 1) study samples had to have a mean age of 11 years old or less 2) did not have any cardiovascular disease, pulmonary diseases (with the exception of asthma), morbid obesity, developmental disabilities, muscular dystrophies, and were otherwise healthy, 3) were not injured. Overweight participants were included in the analysis. Children with asthma were also included as the literature shows they have comparable physical activity levels with children without asthma [15].
Intervention
Regardless of the intervention seen in many of the original articles, only pre-intervention data were used.
Comparator
V̇O2max and ̇V̇O2peak metrics were compared. In addition, we compared studies based on the year and location (USA versus non-USA countries).
Outcomes
The main outcomes were ̇V̇O2max and ̇V̇O2peak metrics measured with treadmill or cycle ergometer incremental testing.
Study design
Articles were considered for the analysis if: 1) they were published in a peer-reviewed journal, 2) mean/standard deviation ̇V̇O2max /̇V̇O2peak parameters was reported, along with mean age data for the subjects, 3) maximal effort was achieved during the incremental test.
These measures resulted with article data on children (girls and boys) that used various testing methods for measuring CRF. Next, all articles reporting CRF values of both sexes combined were excluded from further analysis. The aim was to create sex specific values, that could be later compared.
In addition, analysis excluded articles with graphical results only, field studies, or other non-standardized protocols and types of incremental tests.
A systematic electronic literature search was conducted in PubMed database until 2019 using key search words ((children) AND ((oxygen consumption) OR (aerobic power) OR (peak oxygen consumption) OR (̇VO2) OR (̇VO2max) OR (̇VO2peak)). During the first search, potential articled included boys and girls, that performed incremental tests with cycle ergometry and treadmills. All potential articles up to 2019 were independently hand searched by two researchers. In 2022, the same criteria were used to conduct an additional search from 1. January 2019 to 31. March 2022 (Fig. 1). PRISMA guidelines were used (Additional file 1 2, Figure S1). Data on prepubertal boys using cycle ergometry have been published and presented with different analysis elsewhere [16]. In this meta-analysis, the focus was to explore any differences in aerobic power based on sex, measurement tool or if maximal or peak oxygen consumption metrics are used.
Statistical analysis
Bayesian statistics was used to analyze the results. Bayesian methodology offers a modern, more flexible, and intuitive alternative to frequentists statistics [17–20]. All analyses were conducted through Stan – a state-of-the-art platform for executing modern Bayesian statistical analyses [19]. In this work we were mainly interested how various variables are correlated with measurements of relative VO2. Thus, our response (dependent) variable was the VO2 measurement, while the following variables were explanatory (independent):
-
mean age of study participants,
-
mean body mass of study participants,
-
year of the study,
-
measurement metric (peak or max methodology),
-
country of participants (USA or other),
-
sex (male or female),
-
device (cycle or treadmill).
To analyze how VO2 correlates with mentioned explanatory variables, we used Bayesian multiple linear regression:
$$y | \alpha , \beta , \sigma \sim Normal\left(\alpha +\beta X, \sigma \right),$$
\(\) \(\beta \sim Cauchy\left(0, 2.5\right).\)
We put the standard weakly informative prior on the regression’s beta coefficients (\(\beta\)) [19], we did not put any prior on the intercept (\(\alpha\)) meaning that the default, non-informative prior was used in this case. \(X\) denotes a matrix containing values of explanatory variables, \(y\) is a vector containing values for the response variable and \(\sigma\) is the variance parameter.
In the second analysis, where we analyzed how age along with sex and device influences the probability of measuring the highest (max) or lowest (min) VO2, we fit the above model independently for each of the sex/device combinations (boys on a cycle, boys on a treadmill, girls on a cycle, girls on a treadmill) where age was the only explanatory variable. We then used the model to calculate mean VO2 values as the age of participants changes and compared those to get the reported probabilities.
In all analyses, we used the posterior parameters of our models to estimate certainty of our claims. With Bayesian statistics we can directly quantify the probability (P) of a particular research question, which arguably gives us the most direct and intuitive measure of how certain we are about a claim we are making [17, 18, 20]. In this work we are reporting P as a percentage, meaning that the value of 100 represents the highest possible certainty, labeled as P ≈ 100%. As we work with probabilities in percentages, 100 - P represents the probability that the opposite of a particular claim is true. Uncertainty in our analyses is reported with the Monte Carlo standard error (MCSE) measure.
Ethics
This is a review of available data in databases. Institutional Ethics Committee has confirmed that no ethical approval is required.