The main finding of this study was that the EB test can be used to detect a change in VO2max over a time span of 5 to 8 years. There was a strong and significant correlation between changes in measured and changes in estimated VO2max values (r = 0.75, p < 0.05). Furthermore, the estimation errors (i.e., the difference between measured and estimated VO2max) seems to be rather constant over time (r = 0.84, p < 0.05). This means that anyone being overestimated at the first test, will also be so at the second test. As the measurement error (misclassification) was relatively constant over time, individual characteristics rather than temporary and/or environmental factors seem to contribute to the error.
Previous studies that have investigated whether longitudinal changes in VO2max can be accurately detected via submaximal cycle ergometer tests are scarce. In one study by Åstrand et al., all subjects demonstrated significantly lower values of measured VO2max in 1970 than in 1949, with a mean decrease of 22% in women and 20% in men. These changes was not possible to detect accurately with the submaximal Åstrand test . The mean change in maximal HR was − 15 bpm for women and − 12 bpm for men, whereas some subjects had a rather unchanged value, and others displayed as much as 25–30 beats lower maximal HR. The mean HR at a given submaximal workload was higher at follow-up, but with large inter-individual differences. The explanation for these differences in submaximal HR responses could not be related to differences in VO2max (i.e., there was no correlation between decrease in maximal HR and decline in VO2max) or maximal HR (i.e., there was no correlation between change in maximal HR and HR at submaximal work rate. Consequently, it was not possible to accurately determine the change in measured VO2max through the submaximal Åstrand test and the associated nomograms [9, 23]. In contrast to the present study, those subjects were more homogenous regarding sex and physical fitness, and there was a considerably longer period between tests (the subjects were between 20 and 33 years old at baseline, and between 41 and 54 years old at follow-up).
The estimations of VO2max from submaximal exercise tests are often based on the rather linear relationship between HR and VO2, as well as HR and workload, with increasing exercise intensity. Cycling is a commonly used activity for submaximal exercise testing, since this work mode has relatively small inter-individual variations in work efficiency and because the mechanical efficiency is rather constant in mixed populations, even in populations that have never used or even seen a bicycle or cycle ergometer before . Still, the assumption of a relatively static work efficiency (energy cost/VO2 consumption for produced W) has been questioned. It has been shown that higher body mass results in a higher energy consumption for a given rate of work . In the previously mentioned study by Åstrand et al., the subjects displayed almost no differences in efficiency (measured as O2 uptake for a certain rate of work) in 1949 and 1970. The only exception from this observation was for the work rate of 200 W, where the men had a higher VO2 in 1970 compared to 1949 (2.84 L/min and 2.67 L/min, respectively). However, the sample comprised of healthy, physically fit and normal weight individuals, who had the same body mass on both test occasions. In a subsample in the present study, the VO2 uptake at the standard work rate was unchanged, and both changes in body mass and VO2 at the standard work rate were unrelated to the change in the estimation error, indicating similar work efficiency (data not shown).
The underpinning theory of HR-based exercise tests is often that a decrease in HR at a given submaximal workload is related to a positive training adaptation. A submaximal exercise test may be an efficient and easily accessible method to assess cardiac autonomic activity and to track changes in maximal aerobic capacity [26,27,28]. However, several factors contribute to the overall uncertainty in the interpretation of the HR response to a submaximal workload, and thereby contributes to the misclassification of VO2max from submaximal tests. Several of these factors are of extra importance for the measurement error in a test where the results are derived from the steady state HR at one single work rate, for example the Åstrand test. Environmental conditions, dehydration, ingestion of food and beverage and emotional stress may have an impact on a single HR measurement, as well may different medications alter the HR response to submaximal exercise. For example, treatment with chemotherapy induces autonomic dysfunctions  which may influence HR response to submaximal work, especially at lower levels of exercise . Also in the normal population, it has been shown that intra-individual variations in submaximal HR is greater at lower exercise intensities than at higher intensities . In contrast to the Åstrand test, the EB test is a two-point test (HR measurements at two work rates) with a prediction equation that includes the variable ΔHR/ΔPO (slope). This may counteract some of the measurement errors caused by the above-mentioned external factors and variations in HR response to a given rate of work, attenuating any misclassification at both a single test and at repeated tests.
The variations in VO2max over time can be due to naturally occurring factors such as the consequence of aging, as well as changes in exercise habits and other health-related aspects. Since submaximal tests generally are used in larger-scale settings (even epidemiological studies) it is important to study the ability to capture longitudinal changes (>5 years) in VO2max. However, this is not quite the same thing as experimentally increasing VO2max during a short period of intense aerobic training since the former type of study must take age-related changes into account. The ability of the test to detect training induced increases in VO2max was not investigated in the present study. Stroke volume increases as a consequence of aerobic training and the increased oxygen transport is the primary explanation to increased VO2max, so it is reasonable to hypothesise that the difference in HR between two work rates will be less pronounced. Furthermore, an altered arterio-venous O2 difference (avO2diff), either because of training or aging, may also alter the relation between VO2 and HR.
Regarding the choice of individual high work rate for repeated tests, p-values revealed a significant change in estimation error for the group that was tested with the same work rate at baseline and follow-up tests. However, no difference was found between groups with repeated measures ANOVA. These findings indicate that there is no need to keep the same higher work rate between tests to procure a good estimation. As part of the testing procedure, subjects were asked about their current physical status to select a correct higher work rate, resulting in both lowered and increased load compared to the baseline test. However, there was no significant difference between groups and there is no support to the recommendation to use the same individually chosen higher work rate when monitoring an individual over time. Every test and measurement can be assessed and judged for the specific test occasion. This approach is preferable with repeated tests over time (≥ 5 years), where the subject exhibits a change in physical capacity and consequently needs another individually chosen high work rate.
It has been suggested that the inclusion of cardiorespiratory fitness measurement in routine clinical practice is mandatory to provide an optimal approach for stratifying patients according to risk . Submaximal exercise tests are highly important to provide VO2max values for risk prediction when direct measurements are not feasible. The development of valid and reliable submaximal cycle test, for example the EB test, has high relevance for coaches, researchers and clinicians. The present study showed that there was no significant change in the estimation error for subjects with a decrease in VO2max over time, which allows follow-up measurements to detect a decline in cardiorespiratory fitness and implementation of physical activity and training. However, evaluating long term changes in clinical populations or for training induced changes should be done with caution as this needs to be studied further.
This study has several limitations. The relatively small sample size made it impossible to conduct more detailed analyses in different subgroups, for example individuals with a pronounced change in body mass and/or body composition. It is important to examine the influence of these aspects since the test may be used in association with a weight loss- or training programme. The sample size may also limit the utilisation of BA-plots as a sample size of approximately 50 subjects is recommended in these analyses . However, as this is not an exact requirement, and we have no extreme outliers in or sample, we believe our results to be reliable. A common issue in studies involving maximal testing is the selection bias, often including more fit individuals than in the general population. In the recruitment phase of the present study, individuals with a self-reported currently high training status were more willing to participate in the follow-up tests than those with a self-reported reduction in fitness level. The preponderance of well-trained subjects may eventually limit the accuracy of the prediction equation in the general population, where the normal physiological response to aging is a lowered VO2max.