Interchangeability and optimization of heart rate methods for estimating oxygen uptake in ergometer cycling, level treadmill walking and running

Background The heart rate (HR) method enables estimating oxygen uptake (V̇O2) in physical activities. However, there is a paucity in knowledge about the interchangeability of this method when applied to cycling, walking and running. Furthermore, with the aim of optimization, there is a need to compare different models for establishing HR-V̇O2 relationships. Methods Twenty-four physically active individuals (12 males and 12 females) participated. For each participant, two models of HR-V̇O2 relationships were individually established in ergometer cycling, level treadmill walking and running. Model 1 consisted of five submaximal workloads, whereas model 2 included also a maximal workload. Linear regression equations were used to estimate V̇O2 at seven intensity levels ranging between 25 and 85% of heart rate reserve (HRR). The estimated V̇O2 levels were compared between the exercise modalities and models, as well as with data from a previous study. Results A high level of resemblance in estimated V̇O2 was noted between running and cycling as well as between running and walking, with both model 1 and model 2. When comparing walking and cycling, the V̇O2 levels for given intensities of %HRR were frequently slightly higher in walking with both models (range of significant differences: 5–12%). The variations of the estimated individual V̇O2 values were reduced when using model 2 compared to model 1, both between and within the exercise modalities. Conclusion The HR method is optimized by more workloads and wider ranges. This leads to overall high levels of interchangeability when HR methods are applied in ergometer cycling, level treadmill walking and running. Supplementary Information The online version contains supplementary material available at 10.1186/s12874-022-01524-w.


Introduction
Applying heart rate (HR) methods for estimating oxygen uptake (VȮ2) during physical activities can be of value for education, health promotion and disease prevention as well as research purposes. For these purposes, the HR methods need to be methodologically developed. In this respect, a previous study (1) examined the interchangeability of HR methods between ergometer cycling and level treadmill walking. It found that the individual variation, when applying the HR methods, was reduced when the HR-VȮ2 relationships were based on both submaximal and maximal workloads as compared to only submaximal. In this analysis, we will instead focus on evaluating the effects of using more submaximal workloads (five vs three) as well as wider measurement ranges when establishing the HR-VȮ2 relationships. This is done through comparing results between the present study and Olsson, Salier Eriksson (1).

Methods
Systematic comparisons of HR methods applied to ergometer cycling and level treadmill walking have been made between the present study and a previous one (1). In the present study, 24 participants (12 males and 12 females; age 29 ± 8 years) were included, whereas 34 participants (17 males and 17 females; age 44 ± 6 years) participated in Olsson, Salier Eriksson (1). For further information of the different groups of participants, see the Methods sections in the present study and in Olsson, Salier Eriksson (1).
In both studies, two different models of establishing HR-VȮ2 relationships for cycling and walking, respectively, were examined. Model 1 consisted of only submaximal workloads, while model 2 included also a workload of maximal exercise. Three submaximal workloads were used in Olsson, Salier Eriksson (1), whereas the present study included five submaximal workloads, and a clearly wider range for walking. The different HR and VȮ2 measurement ranges are illustrated in Fig. S1. For further procedures and materials used, see the Methods sections in the present study and in Olsson, Salier Eriksson (1).
To perform this comparative analysis, data in Olsson, Salier Eriksson (1) have been recalculated to be comparable with the present study. Furthermore, all illustrations of individual VȮ2 data in Olsson, Salier Eriksson (1), have been redesigned to be comparable with the figures in the present study (cf. Fig. S2-S5).

Optimization of interchangeability between the exercise modalities
The present comparisons of the estimated individual VȮ2 values of walking vs cycling indicate that there is a reduced individual variation in both models compared to Olsson, Salier Eriksson (1) ( Table  S6; Fig. S2 and S3). This is based on numerically: (1) higher r 2 -coefficients, and (2) narrower 95% confidence intervals for both y-intercepts and slopes in the present study. At the same time, the yintercepts were numerically closer to y = 0 in Olsson, Salier Eriksson (1) than in the present study. Finally, only minor differences were observed when comparing the slopes between the two studies, and they were all close to the lines of identity (Table S6).  Table S6.  Table S6. (1) (n = 34) and the present study (n = 24), ranging between 25-85 %HRR (cf. Fig. S2 and S3). Walk = walking and Cyc = cycling.

Optimization within each exercise modality
In the model comparison of individual VȮ2 values, a high level of resemblance along the line of identity was observed for cycling in Olsson, Salier Eriksson (1), and it did not improve in the present study (Table  S7; cf. Fig. S4a and S5a). On the other hand, the corresponding model comparison for walking showed a higher resemblance in the present study compared to Olsson, Salier Eriksson (1) (cf. Fig. S4b and S5b). This was indicated by numerically; (1) higher r 2 -coefficient, (2) y-intercept closer to y = 0, (3) slope closer to the line of identity, and (4) narrower 95% confidence intervals for both y-intercept and slope (Table S7).  Table S7.  Table S7.  Fig. S4 and S5).

Discussion and Conclusion
There were indications of a reduced individual variation in the exercise mode comparisons of walking vs cycling (Table S6; Fig. S2 and S3) when additional submaximal workloads (five vs three) were used, as well as a wider measurement range for walking. We interpret that this reduction of individual variation enhanced the interchangeability of the HR methods by becoming more accurate.
Given the model comparisons of individual VȮ2 values in cycling and walking, respectively, (Table S7; Fig. S4 and S5), and the fact that the submaximal measurement range for walking in Olsson, Salier Eriksson (1) was almost doubled in the present study (Fig. S1b), there are clear indications that the width of the range is important for the stability of the HR method. Moreover, this indicates that the individual variations between model 1 and model 2 will be reduced when using a wide submaximal range. This decreases the importance of using a maximal workload when establishing HR-VȮ2 relationships and applying HR methods within the same form of exercise.
Finally, note that these comparative analyses of the two studies are based on pairwise numerical comparisons. Due to the low number (two) of possible comparisons, there has been no basis for statistical analyses. Furthermore, the analyses are dependent on two different samples of participants. Therefore we cannot rule out that the observed differences may, at least partly, be due to varying group characteristics rather than to the differences in the HR methods applied. Thus, it would be valuable to also evaluate the issues of optimization strategies through intra-individual comparisons. However, at this stage of analysis, we conclude that more submaximal workloads (five vs three) and wider ranges stabilize the HR-VȮ2 relationships and thereby improve the interchangeability between cycling and walking, as well as optimize the HR methods.