Automatic adjustment of cycle ergometer power output to accurately clamp heart rate

ABSTRACT

We assessed the accuracy and inter-sessional reliability of traditional (manual) compared to automatic (AutoHR) heart rate (HR) clamping methods during submaximal intensity continuous cycling. On separate occasions, thirteen males cycled at an HR corresponding to 80% of the ventilatory threshold for 18 min. Cycling power output was adjusted using either manual or AutoHR methods, encompassing three trials per method. For the manual method, cycling power output was adjusted every 30 s by 0, 5 or 10 W at the experimenter’s discretion. Conversely, AutoHR automatically adjusted power output based on the difference between target and actual HR. Participants’ HR was measured at 1 Hz. Root-mean square error (RMSE) and intraclass correlation coefficients (ICC) were calculated from the difference between measured and target HR to represent accuracy and reliability of each method. The RMSE for the manual method (3.2 ± 2.6 bpm) was significantly higher compared to AutoHR (2.8 ± 2.3 bpm) (p < 0.01, r = 0.13); inter-day ICC were 0.92 and 0.89 for manual adjustment and AutoHR, respectively. Automatic methods to clamp HR are more accurate than manual approaches during submaximal intensity continuous cycling and can be easily implemented for uniform HR control in individual and group training sessions at minimal cost.

KEYWORDS:

• Training loads
• exercise prescription
• internal load
• clamped heart rate

Introduction

Robust sport and exercise science research often involves manipulating exercise intensity to investigate the effects of exercise interventions on clinical and/or performance outcomes across various populations (Impellizzeri et al., 2019). Exercise intensity is commonly assessed using external and internal load metrics. External load metrics, such as power output (PO), indicate the work performed, while internal load metrics, such as heart rate (HR), represent the physiological response to sustained effort (Foster et al., 2017). Many factors (e.g., training status, health, nutrition, environment) likely impact an individual’s response to exercise. Consequently, regulating the individual relative exercise intensity through internal load metrics such as HR would be helpful for applying a consistent stimulus across participants (Impellizzeri et al., 2019). However, controlling exercise intensity via internal load poses a challenge because the external load needs continuous adjustment to match the target intensity, as exercise duration can alter the internal response to a constant external workload. For example, cardiovascular drift, where HR increases after 10 min of moderate intensity exercise, has been well established (Ekelund, 1967). Therefore, it can be challenging to reach and accurately maintain a pre-planned internal response during exercise.

To prescribe exercise intensity using HR, a target HR is initially set, and the external work being performed is adjusted to match the desired HR response. The predominant method for controlling HR during exercise testing currently involves manually adjusting the participant’s mechanical output. For instance, in prior research, investigators manually modified cycling PO every 30 s to maintain an HR corresponding to 60% of maximal oxygen uptake, ensuring a consistent exercise intensity across various environmental conditions (Racinais et al., 2021). However, this traditional method of clamping HR has limitations. Firstly, the magnitude of PO adjustment depends on the discretion of the investigators, which may be affected by inter-rater reliability, particularly with multiple experimenters. Another limitation is the frequency of cycling PO adjustments. Since HR can change within each 30-s interval, significant HR fluctuations may occur during this time, contributing to variability in training outcomes over time. Furthermore, the manual method requires a practitioner to continuously adjust the individual’s PO throughout the session, which may be impractical for large groups performing the exercise. As such, there is a need for a continuous and robust approach to adjusting external load to clamp HR more precisely in practical settings.

A possible solution for this issue is implementing an automatic HR clamp. With this approach, once the target HR is set, a computer programme instantly adjusts cycling PO to enable participants to attain and maintain the planned HR response. For instance, Kawada et al. (1999) used a servo-controller to automatically clamp HR in healthy individuals cycling at 60% and 75% of their HR_max. Their results showed that the time taken to reach 90% of target HR from rest was ~136 s, with a difference of ~3 bpm between the target and actual HR during exercise. Consequently, their findings suggest that an automatic HR clamp may facilitate precise HR regulation. More recently, Li et al. (2023b) maintained HR across hypoxic conditions (2500–4000 m simulated altitude using a custom-built app known as AutoHR. Using this method, cycling PO decreased ~12.3% per 500 m increment in simulated altitude to maintain a given HR, which was achieved without any subjective alterations in exercise intensity from the participant or researcher. Accordingly, tools such as AutoHR can be useful to control relative exercise intensity when a consistent internal load is required over time.

It appears that tools such as AutoHR may offer greater practicality when compared to manual HR clamping methods. Once the target HR is set, no other human inputs are required to sustain the exercise intensity, making HR clamping easily applicable in a group setting. Another advantage of the automatic HR clamp over traditional manual methods is its capacity for real-time reactive adjustments based on the difference between target and actual HR. This feature may result in reduced HR fluctuation during exercise as compared to manual adjustment of cycling PO. Additionally, the use of an algorithm for HR clamping eliminates concerns related to inter-rater reliability. Consequently, tools such as AutoHR appear to be promising for tightly regulating HR during submaximal-intensity exercise. However, to date, there have been no direct comparisons between manual and automatic HR clamping methods.

Accordingly, the aim of this study was to compare the accuracy and inter-session reliability of traditional (manual) versus automatic (AutoHR) methods for HR clamping over three sessions of submaximal intensity continuous cycling for each method. We hypothesised that AutoHR would be more reliable and accurate at maintaining HR during submaximal intensity cycling than the manual method of HR clamping.

Methods

Participants

Thirteen males (age: 26.4 ± 4.3 years; stature: 174.8 ± 6.7 cm; body mass: 74.6 ± 6.4 kg; maximal oxygen uptake: 47.2 ± 5.0 ml·min⁻¹·kg⁻¹), classified as “trained/developmental” athletes (Tier 2; McKay et al., 2021), training a minimum of 4.5 h per week, were recruited. A power calculation for a paired samples t-test (α = 0.05, 1-β = 0.95) was conducted using G*POWER (version 3.1.9.3) to determine the sample size, which suggested a sample size of 10 for our primary variable: heart rate error. The effect size (Cohen’s d) was 1.29, and the critical t-value of 2.26 was determined from previous works (Hunt & Hurni, 2019). Ethical approval was obtained from the Human Research Ethics Committee at The University of Western Australia (ROAP2023/ET000237) in accordance with the Declaration of Helsinki. Participants provided informed consent before inclusion.

Experimental design

This study used a randomised cross-over design. Target HR during cycling sessions was set at 80% of the first ventilatory threshold (VT) determined during baseline testing (Li et al., 2023b).

Individual VT was determined using minute ventilation (V_E), oxygen consumption (VO₂), and volume of expired carbon dioxide (VCO₂) during the incremental cycling test. Specifically, the incremental cycling test involved a ramped increase of external workload at 25 W·min⁻¹ until exhaustion, where the increase in V_E/VO₂ with no increase in V_E/VCO₂ and departure from the linearity of V_E was used to determine VT. Six experimental trials were then conducted (at least 48 h apart) in a randomised order, with HR clamped using either manual or automatic methods; a total of three trials per method were undertaken. Each trial began with a 5-min standardised warm-up, performed at a self-selected intensity rated as 5 out of 10 on the Borg CR10 scale (Borg, 1998). After 5 min of rest, participants cycled for 3 min at an individualised cycling PO corresponding to 80% of VT. Subsequently, they cycled for 15 min at an exercise intensity that maintained their HR at 80% of VT, achieved through either manual or automatic methods.

Manual adjustments

A single experimenter performed manual adjustments to cycling PO for all sessions. During each session, the experimenter monitored actual HR at 1 Hz using a chest strap monitor (Polar H10, Kempele, Finland) displayed on an iPod Touch (Apple Inc., Cupertino, CA, USA). Cycling PO was adjusted at the experimenter’s discretion every 30 s, either by 0, 5, or 15 W to achieve the target HR.

Automatic heart rate clamp

The automatic adjustment of cycling PO to maintain HR was executed through a custom-written application (AutoHR), which was installed on an iPod Touch (Apple Inc., Cupertino, CA, USA) (Li et al., 2023a). The AutoHR app received HR measurements from a chest strap monitor (Polar H10, Kempele, Finland) and used a standardised Bluetooth low energy profile to control the resistance of the cycle ergometer (Wahoo kickR, power trainer v5, Wahoo fitness Inc., Atlanta, GA, USA). To maintain target HR, the AutoHR app employed a series of calculations and control mechanisms. Firstly, it calculated the HR error, defined as the difference between the current and the target HR. Subsequently, it computed the target wattage based on the HR error. The AutoHR app then adjusted the resistance of the ergometer in 5-s intervals using a proportional-integral controller, following the principles of Kawada et al. (1999). Specifically, the proportional controller was defined by the equation: u(t) = Kpe(t) + Ki∫e(t)dt, where Kp and Ki are the proportional and integral gains, respectively; e(t) is the error at time t defined as the difference between the measured HR from the chest strap monitor and target HR set on the AutoHR app. The term ∫e(t)d represents the integral of the error term with respect to time (Aström and Hägglund, 2003). From this equation, the AutoHR app adjusted the ergometer’s resistance as required to match the measured HR with the target HR.

Exercise metrics

During the exercise sessions, AutoHR recorded both cycling PO and HR at 1 Hz for all trials. Average values were then calculated over the last 15 min of cycling for each session.

Statistical analysis

All data are expressed as mean ± standard deviation (±SD). To assess HR clamping accuracy, the root-mean square error (RMSE) between the target and actual HR was averaged per session. Bland-Altman plots with adjusted 95% limits of agreement for repeated measures were plotted (Parker et al., 2020), with the target HR on the x-axis, and the difference between target and actual HR on the y-axis. Intraclass correlation coefficients (ICC) with 95% confidence intervals were calculated using a two-way mixed effects model. A single-fixed rater ICC were used to assess inter-day reliability, with <0.5, between 0.5 and 0.75, between 0.75 and 0.9, and >0.9 indicating poor, moderate, good, and excellent reliability, respectively (Koo & Li, 2016). A paired-samples Wilcoxon test using the R (Team RC. R Core Team, 2024) package rstatix was used to analyse the effects of the adjustment method on RMSE. Additionally, Pearson’s r was calculated for effect sizes (r ≥0.10, r ≥0.30, and r ≥0.50 representing small, medium, and large effects, respectively) (Cohen, 1992).

Results

Mean target HR was 123.3 ± 6.1 bpm (range 115–130 bpm). Mean measured HR were 124.1 ± 7.3 and 123.6 ± 7.1 bpm for manual and AutoHR, respectively. The RMSE for manual control (3.1 ± 2.6 bpm) was significantly higher than for AutoHR (2.8 ± 2.3 bpm) (p < 0.01, r = 0.13). The repeated-measures Bland-Altman plots showed that the bias for manual control and AutoHR were 1.3 and 0.7 bpm, with limits of agreement at 4.7 to −2.2 and 3.4 to −2.1 bpm, respectively (Figure 1).

Figure 1. Bland-Altman plots of AutoHR (a) and manual (b) clamping of heart rate.

Mean difference between target and measured heart rate is plotted as a solid line, along with 95% limits of agreement as dotted lines (±2SD).

AutoHR and manual methods demonstrated excellent and good inter-day reliability ICC (0.92 and 0.89, respectively, all p < 0.01) using the single-fixed rater model. The 95% confidence intervals were 0.97 to 0.81 and 0.96 to 0.74 for AutoHR and manual methods, respectively.

Discussion

This study compared the accuracy and intersession reliability of traditional (manual) versus automatic (AutoHR) methods for clamping HR over multiple sessions of submaximal intensity continuous cycling. Our main findings support our initial hypothesis, indicating that AutoHR is a more accurate method to clamp HR during submaximal intensity cycling. These results align with previous research using automatic HR clamping methods. For instance, in a study maintaining HR during moderate-intensity (66% of HR_max) cycle ergometry exercise, investigators employing comparable algorithms reported a RMSE of 2.6 bpm over 800 s, compared to our results of 2.8 bpm (Hunt & Hurni, 2019). Overall, our findings suggest that AutoHR is an effective tool to tightly clamp HR during cycle ergometer exercise, surpassing the accuracy of manual methods.

To date, this is the first study to investigate the reliability of HR clamping methods. Both manual and AutoHR clamping methods showed good to excellent inter-day reliability, as reflected in ICC values exceeding 0.89. However, the manual method relied on a single experimenter, introducing the possibility of reduced inter-rater reliability based on individual consistency. Conversely, AutoHR adjusts cycling PO through an algorithm, alleviating concerns related to inter-rater reliability and freeing up the practitioner to focus on other tasks beyond load control. Overall, our findings show that AutoHR and manual methods of clamping HR are reliable for controlling exercise intensity during cycle ergometer exercise.

Practically, AutoHR offers an easily implemented and cost effective method to accurately clamp HR, eliminating the need for constant supervision by a practitioner. This method ensures uniform HR control during both individual and group training sessions. Additionally, across multiple sessions, AutoHR can reliably control cycling PO according to a target HR, ensuring training consistency when using HR to prescribe individual athlete training sessions.

Limitations

A limitation of our investigation is that our reliability results for manual cycling PO adjustments pertain to a single experimenter familiar with the equipment, compared to a generic operator. Therefore, the reliability outcomes may vary when manual adjustment of cycling PO is performed by different experimenters. Nevertheless, our results suggest that AutoHR is a reliable and accurate tool for consistently controlling cycling PO based on a target HR, which removes potential inter-rater differences that may arise from the manual approach.

Conclusion

Automatic HR clamps, such as AutoHR, provide accurate and reliable control of exercise intensity during submaximal cycle ergometer exercise. We anticipate that such approaches will garner increased attention for prescribing and controlling exercise intensities in both research and practical settings, particularly those involving ergometer-based exercises.

Acknowledgments

The authors thank the participants for their dedication, commitment, and cooperation with this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Author BRS is supported by an Investigator Grant from the National Health and Medical Research Council (APP1196462).