Comprehensive monitoring of fitness and performance as well as accurate diagnosis of fatigue, non-functional overreaching, and overtraining states are crucial for optimizing training and reducing risk of injury in elite professional sport [1
]. In this regard, sensitive, non-invasive, time-efficient and cost-effective testing methods and biomarkers encompassing a multidimensional approach are being sought by coaches, exercise scientists and sports physicians to improve the evaluation of athletes [4
Over the past decade, parameters associated with autonomic nervous system (ANS) regulation such as heart rate (HR) and heart rate variability (HRV), measures assessed during the post-exercise recovery period, have received increasing interest for monitoring training status and cardiovascular fitness [2
]. Nevertheless, contradictory findings related to methodological inconsistencies and partial misinterpretation of the results limit the widespread implementation of HR and HRV measures in the sports field [4
From the perspective of coaches and sports professionals, it is important to evaluate training status in as many athletes as possible relatively quickly and frequently to distinguish intended (e.g., due to training) from unintended (measurement error) changes using reliable measurements and validated tools. This will ensure reproducible results and enable meaningful findings [26
]. Parameters (or tools) are rated as useful or sensitive based on providing high reliability and low test–retest variation [4
Previous studies on reliability of HRV measures in athletes focused mainly on looking for an ultra-shortened reliable user-friendly HRV parameter [22
]. It was suggested that the root mean square of successive differences between adjacent normal RR intervals (RMSSD), or its log transformed version (lnRMSSD), calculated based on 60 sec recordings is reliable [30
] and displays strong agreement with RMSSD criterion derived from 5 min recordings [28
]. Moreover, RMSSD is suggested to be the most appropriate and attractive parameter for use in elite endurance and team sports athletes [2
Apart from the cardiac measurements, the basic parameter reflecting respiration (another part of the cardiorespiratory fitness) is respiratory rate. The assessment of its reliability and agreement can also be considered relevant, as some moderate causal effects might also influence the measurements of cardiac parameters.
A less-studied group of endurance athletes are elite modern pentathletes. The modern pentathlon is an Olympic sport that consists of five different modalities (fencing, freestyle swimming, equestrian show jumping, and a combination of pistol shooting and cross-country running). Events last up to 8 h in duration, making energy and physiological demands close to maximal [35
]. To date, there are few studies addressing modern pentathlon athletes [35
] and we have found no data within the literature on the reliability of HRV in this population. Establishing the typical variation in HRV among this population is necessary to aid coaches in detecting meaningful changes related to pentathletes’ training status.
Despite the large number of methodological papers published on HRV [42
], many studies in this area failed to provide the necessary details concerning data acquisition and measurements, so the experimental design could not be replicated in laboratories, clinical settings and sports field. The lack of details concerning methodological aspects of the study significantly limits confidence in interpretation [52
]. A recent paper on HRV reliability stated that still little is known about the reliability of baseline 5 min (short-term) measurements of HRV and studies continue to differ with respect to important methodological characteristics [54
]. Therefore, the presented study has the following aims: (i) to assess the inter-day reliability of short-term (5 min) HR, respiratory rate and selected time-domain, frequency-domain, and non-linear HRV parameters; (ii) to assess the inter-day reliability of ultra-short-term (1 min) HR, respiratory rate and RMSSD (popular in sports field); (iii) to determine the agreement between short-term (5 min) and ultra-short-term (1 min) parameters, and (iv) to verify the correlations between differences in HR, respiratory rate and HRV parameters in stable conditions in elite modern pentathletes.
The purpose of this pilot study was to assess the reliability of short-term (5 min) and ultra-short-term (1 min) HRV parameters derived from ECG recordings performed in stable measurement conditions with one-week time interval between tests in elite modern pentathletes. Agreement between short-term and ultra-short-term parameters and correlation between differences in HR, RespRate and HRV parameters were also assessed.
We showed that short-term (5 min) HR, RespRate, lnRMSSD (time-domain), lnHF (frequency-domain), SD2/SD1 (nonlinear) and ultra-short-term (1 min) HR, RespRate and lnRMSSD presented acceptable, satisfactory reliability. These results indicate that the aforementioned HRV parameters could be used by coaches and researchers as reliable parameters in endurance athletes during baseline periods in laboratory-controlled settings. Methodological differences between published studies, and general lack of studies on test-retest reliability analysis of HR, RespRate, short-term (5 min) and ultra-short-term (1 min) HRV parameters in elite athletes hinder comparisons. The current findings are somewhat in line with previous studies showing that time-domain indices are more reliable than spectral-domain parameters in moderately-trained males [11
] and that ultra-short-term (1 min) lnRMSSD demonstrated acceptable interday reliability (ICC = 0.90, CV < 7%) in elite rugby union players [30
Recently, many authors have adopted ultra-shortened, user-friendly HRV parameters for the evaluation of endurance and team-sport athletes; however, the analysis of the relevance of such parameters focused more on agreement with 5 min criterion parameters than on test-retest reliability [22
]. Ultra-short-term (1 min) lnRMSSD has been shown to present strong agreement with criterion 5 min recordings [28
]. No significant differences between short-term (5 min) and ultra-short-term (1 min) HR, RespRate and lnRMSSD were observed in the current study, which is somewhat in line with previous studies with endurance and team-sport athletes [22
However, in our study, all analyzed ultra-short-term parameters, in both Test and Retest, showed a LoA with short-term parameters that exceeded the defined a priori maximum acceptable difference (SWC). This is contradictory to the high ICC values obtained between 5 min and 1 min HRV parameters. In studies evaluating agreement between short-term and ultra-short-term parameters, authors have not provided the maximum acceptable difference, nor did they report whether it was exceeded by the LoA. Therefore, the criteria for defining strong agreement in these studies are not well defined. Recent reviews have highlighted that important data from the Bland–Altman method are often omitted [79
]. Abu-Arafeh et al. recently provided a comprehensive list of key items for reporting Bland–Altman analysis [79
]. The first key item, not reported in published studies on agreement between ultra-short-term and criterion HRV parameters in athletes, is the definition of the a priori acceptable LoA, to define the minimal agreement needed to consider the new measurement method as interchangeable with another method (often gold standard or criterion method) [79
]. lnRMSSD justifiably seems to be the preferred HRV index for monitoring athletic performance and training status due to several advantages described previously [2
]. We confirm that ultra-short-term (1 min) lnRMSSD, in combination with HR and respiratory rate, can be reliably used in elite modern pentathletes. However, without defining a priori the maximum acceptable difference between 5 min and 1 min parameters, caution should be taken before considering these two methods as interchangeable, at least in this group of participants. From the results of our study, no parameter can be considered interchangeable for 1 min and 5 min, once the LoA between them exceeds the SWC.
On the other hand, although the consideration of LoA < SWC represents an important criterion for agreement analysis, we believe it has an important limitation. Since LoA is calculated from the SD of differences between “Test” and “Retest”, LoA will be low whenever the differences from all subjects tend to be homogeneous. This does not happen only when the results from “Test” and “Retest” are the same, but also when they vary by the same amount in all individuals. In other words, the fixed bias is not taken into account in this comparison. We suggest that a more complete analysis would involve the comparison of LoA and SWC together with the one-sample t-test to check if the fixed bias is different from zero. However, since in our study no parameter presented a LoA wider than SWC, the one-sample t-test was needless.
In experimental conditions similar to those of the current study, the reliability of time-domain and frequency-domain HRV parameters was shown to be affected by differences in HR and RespRate in young healthy volunteers. HR was a stronger determinant for HRV reproducibility than RespRate, and even a minimal change of HR considerably altered HRV [83
]. In elite athletes, the differences in short-term time-domain and nonlinear HRV parameters, as well as in ultra-short-term HRV parameters, turned out to be more correlated with differences in HR than with differences in RespRate. This supports the notion that changes in HRV indices should be assessed and interpreted with regard to concomitant changes in resting HR [4
]. The normalization of HRV parameters concerning the HR level has already been suggested and we reinforce its importance for comparisons of people with different levels of HR [67
Contrastingly, differences in short-term frequency-domain HRV parameters were more correlated with differences in RespRate than with differences in HR. Indeed, HRV (mostly frequency-domain parameters) is also highly affected by respiratory depth and rate [44
]. Spectral powers at LF and HF bands are commonly attributed to sympathetic and vagal, and vagal influences alone on HR, respectively. Thus, coaches and sport practitioners should be aware that changes in the respiratory rate may confound spectral indices [53
]. It has been suggested that depth of breathing could be even more important than its rate [5
]. Młyńczak and Krysztofiak proposed cardiorespiratory temporal causal links with the path for lying supine from tidal volume, through heart activity variation and average heart activity, to respiratory timing [5
]. Even if those links appeared rather moderate, the possible differences in the respiratory pattern should be taken into account in the protocol assessing the cardiac parameters. The measurement of tidal volumes using, e.g., mouthpieces may affect HRV data [17
]. Moreover, there is no optimal solution on how to record and control respiratory depth and rate in HRV studies [44
]. One possible solution for future studies in athletes could be to implement the Pneumonitor 2 or 3, a portable device that would register respiratory pattern (both depth and rate) together with single-lead ECG (enough to estimate aforementioned parameters), motion, and/or pulse oximetry (saturation, pulse wave) designed for environmental physiology analyses and sports medicine [86
Although HRV parameters are often used as biomarkers of physical conditioning, as well as indicators of the severity of diseases, the physiological meaning of such parameters are not well understood in many situations. While the spectral power at the HF band is widely accepted to represent the vagal modulation (respiratory sinus arrhythmia) to the heart under normal respiratory frequencies (>9 breaths/min), the physiological interpretation of other indices is not so clear. This is the case of most nonlinear HRV parameters. For example, the short-term scaling exponent (DFAα1) represents the fractal correlations present in RR series and was demonstrated to represent one of the best risk factors for patients with heart failure or myocardial infarction [88
]. However, the changes in fractal structure of RR series in these patients are likely to be the consequence of the global change in cardiovascular function, and not a marker of any specific physiological variable. The living organism can be considered a complex system and, as such, it is difficult to disentangle the influences caused by each mechanism, once their functions are highly interdependent [90
]. This is the reason why many nonlinear HRV parameters are considered as “complexity measures”, in the sense that, although they are not clearly associated with specific physiological meanings, they are able to represent the general complexity of the system, attested by their important role as risk factors, prognosis and fitness [91
No previous study on inter-day reliability of HRV in athletes with similar methodological characteristics to our study was found. As underlined by many authors, there are still substantial methodological inconsistencies throughout the literature regarding HRV in sports science that limit comparison [4
]. Moreover, the authors of a recent study on test-retest reliability of short-term HRV parameters suggested that the studies differ with respect to important methodological issues and evidence on this topic is far from clear [54
]. There are several methodological characteristics that may cause differences in findings among studies on HRV reliability, e.g., reliability statistics adopted, number of tests performed and the time interval between them, and sample heterogeneity [54
]. Lack of information about other methodological aspects also limits potential for study replication and confidence in interpretation [52
], such as the lack of information on ECG acquisition and processing, including device, software, recording duration and conditions, breathing control, and position during recordings. Thus, we emphasize the importance of providing all the necessary information for reproducibility, as well as the standardization of statistics for the analysis of reliability of biomarker parameters in athletes.
Our pilot study was limited by a small homogenous sample size taken at a given moment of the sport season, which limits generalizability of the study’s findings. The consequence of low sample size is the illustrative character of the correlation data. HRV data were obtained during supine ECG recordings in controlled laboratory settings. Heart rate monitors or smartphone with HRV application may be more useful to collect RR intervals in athletes than traditional ECG, especially outside of a laboratory setting [94
]. Nevertheless, as suggested by Lucini et al., caution must be applied when assessing HRV using devices that cannot discriminate RR series between sinus and non-sinus beats, especially in athletes [57
]. In our study, four participants were excluded from the analysis due to presence of cardiac abnormalities. Recently, abnormal ECG changes were observed in about 4% of top-level endurance athletes [97
]. Therefore, we suggest performing ECG screening among examined athletes before including the data for detailed HRV analysis.