Validity and sensitivity of field tests’ heart-rate recovery assessment in recreational football players

We aimed at examining the criterion validity and sensitivity of heart-rate recovery (HRRec) in profiling cardiorespiratory fitness in male recreational football players in the untrained and trained status, using endurance field-tests. Thirty-two male untrained subjects (age 40 ± 6 years, VO2max 41.7 ± 5.7 ml·kg-1·min-1, body mass 82.7 ± 9.8 kg, stature 173.3 ± 7.4 cm) participated in a 12-week (2‒3 sessions per week) recreational football intervention and were tested pre- and post-intervention (i.e. untrained and trained status). The participants performed three intermittent field tests for aerobic performance assessment, namely Yo-Yo intermittent endurance level 1 (YYIE1) and level 2 (YYIE2) tests, and Yo-Yo intermittent recovery level 1 (YYIR1) test. VO2max was assessed by performing a progressive maximal treadmill test (TT) and maximal HR (HRmax) determined as the maximal value across the testing conditions (i.e., Yo-Yo intermittent tests or TT). HRRec was calculated as the difference between Yo-Yo tests’ HRpeak or HRmax and HR at 30 s (HR30), 60 s (HR60) and 120 s (HR120) and considered as beats·min-1 (absolute) and as % of tests’ HRpeak or HRmax values. Significant post-intervention improvements (p<0.0001) were shown in VO2max (8.6%) and Yo-Yo tests performance (23–35%). Trivial to small (p>0.05) associations were found between VO2max and HRRec (r = -0.05−0.27, p>0.05) across the Yo-Yo tests, and training status either expressed as percentage of HRpeak or HRmax. The results of this study do not support the use of field-test derived HRRec to track cardiorespiratory fitness and training status in adult male recreational football players.


Introduction
Heart rate (HR) monitoring is a valid and popular method for controlling aerobic training aimed at enhancing cardiorespiratory health [1]. In individuals with different training status, health conditions, age and sex, maximal exercise and recovery HR are the variables usually considered to prescribe and monitor training, and to assess cardiorespiratory fitness [1,2]. Heart rate recovery (HR Rec ) is most commonly measured as the rate at which heart rate decreases within the following seconds/minutes after the end of exercise [3,4] and reflects the dynamic balance and coordinated interplay between parasympathetic reactivation and sympathetic withdrawal [4][5][6]. HR Rec following exercise to exhaustion was deemed sensitive to the interplay between parasympathetic and sympathetic nervous activity, reflecting autonomic efficiency [2,3,7]. This, alongside with the high accessibility to HR Rec measurements, promoted the development of normative values considered useful for detecting pernicious variations in post-maximal-exercise HR kinetics in daily practice [3,[8][9][10]. Indeed, faster HR Rec was reported to be associated with a higher fitness level, and subjects with abnormal HR Rec (i.e. a decrease of � 12 beats�min -1 for HR at 60 s after the end of the test) [3,8] were less likely to be engaged in regular and strenuous exercise [8]. Furthermore, HR Rec revealed to be a prognostic indicator of adverse cardiometabolic outcomes and an independent factor for metabolic syndrome prediction [11,12]. The published scientific evidence of a deleterious effect of attenuated HR Rec on cardiovascular and metabolic health and all-cause mortality, promoted HR Rec recording in clinical practice as "per se" routine for health risk assessment [12].
Training interventions using conventional aerobic exercise promoted positive changes in HR Rec in cardiovascular patients and in athletes [13,14]. However, most of the published research studies were carried out considering pre-to post-intervention HR Rec at selected time points post-exercise, as training outcome variables, not considering changes in VO 2max . In fact, the supposed low sensitivity of VO 2max for tracking cardiorespiratory fitness changes, and its low absolute reliability, promoted the consideration of HR Rec for performance monitoring [13]. Nevertheless, VO 2max levels relate to cardiorespiratory health and to the likelihood of allcause or cardiovascular mortality, suggesting consideration of VO 2max tracking in cardiorespiratory fitness enhancing programmes [15]. The practical interest in evaluating cardiorespiratory fitness with an easily accessible variable like HR Rec and with endurance field tests, warrants therefore experimental consideration in recreational sports [16,17].
Recreational football research has provided compelling evidence of clinically sound training-induced improvements in cardiorespiratory fitness and aerobic performance [18,19]. Regular weekly practice of recreational football in the form of small-sided games, has been proposed as an alternative exercise mode for improving cardiovascular health across age, sex and health status. The casually intermittent nature of recreational football and the associated variability of the individual responses to practice (i.e. small-sided games), suggests the periodical evaluation of aerobic fitness to assess the effectiveness of the training programmes [17,20]. Furthermore, recreational football involving high-intensity bouts of exercise interspersed with activities performed at lower intensity for recovery, may constitute a viable training activity for improving HR Rec [18,21].
To the best of these study authors' knowledge, no research has been published with the aim of evaluating the validity and sensitivity (i.e. external responsiveness) of HR Rec in recreational football players. Information about the validity, sensitivity and applicability of HR Rec monitoring in recreational football would be of great practical importance for the control, regulation and implementation of successful training programmes.
The main aim of this study was therefore to examine the association between HR Rec values obtained at arbitrarily chosen time points after intermittent maximal field tests with VO 2max in adult recreational football players (convergent construct validity) in the untrained and trained states (i.e. longitudinal construct validity). Recovery HR was assessed using field tests deemed to induce exhaustion and popularly used in recreational football interventions (i.e. Yo-Yo intermittent tests) [17,20]. An effect of individual and training-induced cardiorespiratory fitness improvements on HR Rec was assumed as working hypothesis [13].

Participants
In this study, thirty-two male adults (age 40 ± 6 years, VO 2max 41.74 ± 5.72 ml�kg -1 �min -1 , body mass 82.7 ± 9.8 kg, stature 173.3 ± 7.4 cm, systolic and diastolic blood pressure 125 ± 11 and 74 ± 8 mmHg, respectively) volunteered to participate. The participants were tested at the untrained and trained states, i.e., before and after engaging in a 12-week recreational football training-based intervention. The untrained state (baseline conditions, i.e., pre-intervention) was defined as the participants having less than 20 min of exercise on 3 or more days a week [22]. All the participants were familiarised with the procedures used in the investigation during the two weeks before the commencement of the study by performing submaximal versions of the treadmill test and the Yo-Yo intermittent tests. The participants gave their written informed consent to participate in the study, which was conducted in accordance with the Declaration of Helsinki, and ethical approval was provided by the Ethics Committee of the Faculty of Sport, University of Porto (Porto, Portugal). All participants were informed of the risks and benefits of participating and made aware that they could withdraw from the study at any time without penalty.

Design
In this study, HR Rec was determined as the difference between the Yo-Yo intermittent tests' peak HR (HR peak ) or maximal HR (HR max ), depending on whether HR max or only HR peak was reached during the test conditions, and post-exhaustion HR at selected time points, i.e. 30s (HR30), 60s (HR60) and 120s (HR120) after the end of the tests [3,9]. Specifically, HR Rec (i.e. ΔHR Rec = peak/maximal HR minus post-exhaustion HR value) was reported in absolute values (beats�min -1 ) and as a percentage of HR peak or HR max (%HR Rec ) reached during the tests [23]. With the aim of evaluating the proposed levels of validity, data normalisation was performed using HR peak and HR max in either the untrained or trained states. Maximal HR (HR max ) was assessed as the maximal value reached across the testing conditions (i.e. Yo-Yo intermittent tests or the treadmill test for VO 2max assessment), using a multiple approach, as suggested by Póvoas et al. [16], in recreational football. HR peak refers to the maximal value reached during a testing condition that requires maximal effort, but that is below the maximal reached by the participant in all testing conditions. The magnitude of HR 60 was rated for clinical importance using the cut-off values suggested by Cole et al. [3,24]. Given the relatively active recovery observed post-exhaustion during the field tests (deceleration and spontaneous ambulation), abnormality was considered when HR 60 was �12 beats�min -1 [3,8].
The intensity and duration of the exercise used to induce HR Rec has been considered as a confounding variable [13]. With the aim of examining the interest in using intermittent endurance field tests in assessing HR Rec , three intermittent versions of the Yo-Yo test were considered [17], namely levels 1 and 2 of the Yo-Yo intermittent endurance test (YYIE1 and YYIE2, respectively) and the Yo-Yo intermittent recovery test level 1 (YYIR1). The field test protocols were assumed to induce similar aerobic demands with different anaerobic involvement and time to exhaustion in order to stress different HR Rec [13,17].
After the baseline (i.e. untrained status) VO 2max and field testing, the participants engaged in a recreational football training intervention (2-3 60-min weekly sessions) and were retested after 12 weeks of training to access the responsiveness of the selected variables (i.e., pre-and post-intervention). The training intervention was carried out according to the guidelines suggested by Krustrup et al. [18,19,25] for recreational football interventions with male participants.

Testing procedures
The field tests (Yo-Yo intermittent tests) and the treadmill test for VO 2max (TT) assessment were performed in random order with at least 4 days (i.e., 4-6 days) of recovery in between. Test standardisation was achieved by performing the Yo-Yo intermittent tests on the same artificial football pitch and at the same time of day for circadian performance consistency. Furthermore, a standardised warm-up consisting of 10 min of running at different intensities and with changes of direction preceded each Yo-Yo intermittent test. Two minutes of passive rest were considered for each of the participants, before the start of the field tests. On the day before testing, the players refrained from vigorous physical activity.
The proposed Yo-Yo intermittent tests differ in their initial running speed and progression, and the between-bouts (40 m) recovery lasts 5-10 s, during which the participants are asked to cover 5-10 m. The Yo-Yo intermittent test protocols were implemented according to the procedures suggested by Krustrup et al. [26][27][28].
The TT (HP Cosmos Quasar, Nussdorf, Germany) consisted of 3 min of walking at 5 km�h -1 and 2 min of running at 8 km�h -1 with 0% inclination, and then alternating between increases in speed (1 km�h -1 ) and inclination (1%) every 30 s until voluntary exhaustion. Expired respiratory gas fractions were measured using an open-circuit breath-by-breath automated gas analysis system (Quark CPET, Cosmed, Rome, Italy). Attainment of VO 2max was assumed when the participants achieved a plateau in VO 2 despite an increase in exercise intensity and at least one of the following criteria: a respiratory exchange ratio (RER) greater than 1.10 and RPE equal to or higher than 7 [29,30]. The highest 15-s VO 2 during the final stages of the test was considered as proof of individual VO 2max [16,17]. Data analysis was performed with manual inspection of each TT data file using an Excel file (Microsoft, Redmont, USA).
Attainment of individual maximal effort during field tests was considered when the participants, at their subjective exhaustion, reported a rating of perceived exertion (RPE) equal to or higher than 7, at a 0-10 scale or had a HR peak equal to or higher than 90% of their age-predicted HR max . Visual inspection of HR profile was performed to assess possible artefacts and to evaluate possible HR plateau and peak.
All exercise HRs were recorded at 1-s intervals using Polar Team System 2 HR monitors (Polar Electro Oy, Kempele, Finland). The players were allowed to drink water ad libitum in order to ensure proper hydration under all the exercise conditions considered in this study. After the completion of the field tests participants were instructed and guided to stay with minimal movement to standardise the recovery (2−3 min). No drinking was allowed during the recovery period.

Training intervention
After baseline testing (i.e. laboratory and field testing, n = 32), the participants engaged in a recreational football intervention comprising 2-3 60-min training sessions per week in the form of 45-min small-sided games played on an artificial pitch (7v7; 43 x 27 m pitch, 83 m 2 per player) [31]. The training intervention was conducted over 12 weeks, and the intensity of the sessions was monitored using HR monitors and the subjective internal load estimated by the RPE method [32]. All participants repeated all the test procedures post-intervention in the week after the completion of the last recreational football training session. Participants were advised to follow the guidelines followed at baseline testing.

Statistical analyses
Results are expressed as means±standard deviations (±SD) and 95% confidence intervals (95% CI). Normality assumption was verified using the Shapiro-Wilk W-test. A repeated-measurements analysis of variance (ANOVA) with post-hoc Bonferroni test was used to compare HR Rec across the tests' recovery time points (i.e. HR 30, HR 60, HR 120 ). Practical differences were assessed as partial eta squared (η 2 p ) and magnitudes rated as follows: η 2 p �0.14 large effect, 0.14>η 2 p �0.06 medium effect, 0.06>η 2 p �0.01 small effect and η 2 p <0.01 trivial effect [33]. Pearson correlation (r) was used to assess the associations between variables. The magnitude of the reported effects was described using the Hopkins et al. [34] criteria. Within-test conditions variability was expressed as coefficient of variation (%CV). Relative reliability was assessed using the intraclass correlation coefficient (ICC 3,1 ) with 95% CI [35,36]. According to Landis and Kock [37], ICC values of 0.00-0.20, 0.21-0.40, 0.41-0.60, 0.61-0.80, 0.81-1.00 were considered as slight, fair, moderate, substantial and almost perfect, respectively. The Cohen's d was used to evaluate the effect size, with values above 0.8, between 0.8 and 0.5, between 0.5 and 0.2, and lower than 0.2 considered as large, moderate, small and trivial, respectively [24]. The smallest worthwhile change (SWC) in measurement was considered to test the practical difference between variables and calculated as 0.2 times the variable standard deviation [34]. Sample size estimation was performed for a sample power of 85% with an effect size of 0.50 at a significance level of 5%, resulting in 29 participants to be recruited. Significance was set at 5% (P< 0.05).

Results
The participants showed a relative mean attendance of 73 ± 15% (26 ± 5 total training sessions out off a maximum of 36) with a weekly average of 2.2 ± 0.5 training sessions. The VO 2max (8.6%,~1 MET) and Yo-Yo tests performances were significantly (large) improved (23, 37 and 35% for YYIE1, YYIE2 and YYIR1, respectively) after the training intervention (Table 1). A large and significant decrement (-2%) in HR max was detected after 12 weeks of recreational football training (Table 1). Yo-Yo test HR peak was significantly decreased (3, 1 and 1% for YYIE1, YYIE2 and YYIR1, respectively) post-intervention (small to large). The relative reliability (pre-to post-intervention) of the above variables was substantial to almost perfect (ICC>0.71).
HR values during the considered recovery time points are reported in Table 2 as absolute (beats� min -1 ) and relative (% of tests' HR peak or HR max values). Large and significant (p<0.0001) differences were reported for the test conditions across the selected recovery time points and on the two testing occasions (i.e. pre-and post-training intervention). Participants achieved 87-94, 79-86 and 67-74% of their HR max or HR peak values at HR 30 , HR 60 and HR 120 , respectively. The corresponding absolute HR difference ranges were 11-24, 25-39 and 51-61 beats�min -1 for HR 30 , HR 60 and HR 120 , respectively.
Significantly higher (small) relative (%) post-training HR 30 values were found in YYIE1 when using HR peak or HR max for normalising HR Rec (Table 3). Higher (small, p<0.04) postintervention %HR 30 values in YYIR1 were reported for both HR peak and HR max . Lower and significant (p<0.04) post-intervention %HR 60 values were seen for HR peak (moderate) and HR max (small) in the YYIE2 test. When considering %HR 120 , a significant and moderate decrement was evident in YYIE1 for both HR peak and HR max . The YYIE2%HR 120 was small and significantly lower when considering HR peak for normalisation.
At baseline, VO 2max was not significantly associated (trivial to moderate) with HR Rec at the selected time points, expressed as test HR peak or HR max , in any of the considered field-testing conditions (i.e. YYIE1, YYIE2 and YYIR1, Table 4). The lack of significant (p>0.05) correlations persisted in the trained state.
Using the �12 beats�min -1 criterion to qualitatively evaluate HR Rec , only 3-6% and 3% of the participants reported the supposed abnormalities in HR 60 during the pre-intervention YYIE2 and post-intervention YYIE1, respectively (Table 5).

Discussion
This is the first study to examine the validity and sensitivity of using intermittent endurance field tests' post-exhaustion HR Rec values to characterise cardiorespiratory fitness in male participants that volunteered for a recreational football intervention, in the trained and untrained states [13,17]. These tests were supposed to induce similar maximal aerobic demands and different anaerobic loads [38,39]. The main finding was that no associations of significance or practical importance were found between HR Rec and cardiorespiratory fitness in either the untrained or trained states. The reported significant, mainly moderate to large, changes in post-intervention HR peak and HR max affected the representation of HR Rec values magnitude, Table 2

. Heart rate (HR) recovery values at selected recovery time points after the field tests and between their values differences pre-to post-training intervention.
The HR values are reported as % of test peak HR (HR peak ) and individual maximal HR (HR max ).  In this study, the relative reliability of VO 2max , HR max and Yo-Yo HR peak pre-to post-intervention was almost perfect (ICC, 0.81-0.93), supporting the consistency of participants' training-induced changes ranking [37]. Interestingly, significant decrements in HR max (large) and Yo-Yo HR peak (small-to-large) were found at post-intervention (Table 1) [40]. These results provide evidence of the internal validity of this study design and confirm the applicability of the Yo-Yo intermittent tests in recreational football [16,17,20].

Table 3. Heart rate (HR) recovery values (%) at selected time points before (pre) and after (post) the training intervention and considering test peak HR (HR peak ) and across tests maximal HR (HR max ), as normalizing variables.
Bosquet et al. [23] assessed HR Rec reliability by replicating a maximal treadmill test at least 72 hours apart in healthy subjects. A general low relative reliability of ΔHR Rec variables was reported, except when considering absolute HR Rec values (beats�min -1 ), with ICC ranging between 0.68 and 0.83 [23]. In line with the cited study, we found a substantial agreement between pre-to-post absolute HR Rec values (0.67-0.79) across the post-exhaustion time points [23]. Similarly, mainly slight to moderate agreement (0.17-0.61 and 0.05-0.63 using HR peak and HR max , respectively) was reported for ΔHR Rec across the Yo-Yo tests and recovery time points. These results (i.e. long-term relative reliability) add evidence of the poor reliability of ΔHR Rec values when used to characterise HR Rec after maximal testing [23]. This finding is of practical relevance as, in the clinical set-up, HR Rec efficiency is mainly reported as ΔHR Rec [3,14,41].
HR Rec is deemed to be affected by the fitness level and training status of subjects of different age, gender and health condition [5,7,13,42]. Cross-sectional studies found cardiorespiratory fitness as a possible cause for the faster HR Rec in athletic populations compared to untrained or inactive subjects [13]. Darr et al. [41] reported an effect of VO 2max level on HR Rec with trained subjects (>60 ml�kg -1 �min -1 mean VO 2peak ), reporting a faster decrement in HR than in untrained subjects. Interestingly, the untrained male subjects' mean VO 2max values (40 ml�kg -1 �min -1 ) were similar to those of this study's recreational football participants at preintervention evaluations (i.e. untrained status), suggesting an effect of training status on HR Rec . This study's findings did not provide empirical evidence of an association between cardiorespiratory fitness improvements and HR Rec variations in recreational football players. This was supported by the analyses performed in both the untrained and trained status and occurs Table 5. Frequency (count and %) of participants with heart rate (HR) recovery difference �12 beats�min -1 in the untrained and trained state (i.e., at pre-and postintervention). irrespective of consideration for physiological (i.e. VO 2max ) or performance (Yo-Yo tests) changes. Additionally, this study's results are in line with those of Hautala et al. [43] in inactive subjects who trained for 2 weeks on an intensive endurance programme and reported no variations in HR 60 , despite significant positive VO 2max changes (+8%). Similarly, 3 weeks of intensive football training did not provide any changes in HR Rec in competitive young football players [44]. Again, comparison with other studies addressing HR Rec may be confounding, as different research designs and HR Rec assessment tests were used [45]. The reported large and significant reduction in HR max and HR peak across the Yo-Yo tests further supports the previously reported findings of training-induced changes in heart physiology [40]. Although not based on robust mechanistic evidence, the reported significant and practical important decrement in HR max could have been the result of variations in short-term neurological changes related to variation in parasympathetic and sympathetic nervous systems interplay [40]. Short-term effects of recreational football on players' heart anatomy and physiology, may have also played a role [46]. However, this recreational football intervention revealed a limited effect on HR Rec , with moderate to large changes in absolute HR Rec values observed only for HR 120 after YYIE2 and YYIE1 and for HR 30 and HR 60 after YYIE2 (Table 1), suggesting a test-dependent effect on HR Rec . This was further supported by ΔHR Rec analyses, revealing significantly higher post-intervention values only for a few test conditions, i.e. HR 60 in YYIE2 and HR 120 in both YYIE2 and YYIE1, when considering HR peak as a reference for normalisation. Interestingly, a faster post-training intervention HR Rec was only evident for the YYIE1 ΔHR 120 when considering HR max . The variations in HR Rec across the tests and testing time points were also evident when considering %HR Rec values with a general increase (trivial to moderate) in post-intervention values.
Improvements in VO 2max and aerobic performance as a consequence of endurance training and recreational football practice were shown to be associated with significant decrements in HR max and changes in HR peak in field tests [16,20,40]. In this study, large and significant decrements in HR max were detected, providing practical relevance of changes in the range of 3-6 beats�min -1 (SWC 2 beats�min -1 ). Given the period of the considered change (12 weeks), a reassessment of HR max every 12 weeks may be advisable for controlling and regulating exercise intensity in recreational football interventions. The reported changes in HR max were paralleled by moderate to large decrements (p<0.04) in Yo-Yo HR peak post-intervention, further suggesting caution when choosing ΔHR Rec to profile HR Rec . Indeed, variations in peak HR values as an effect of training, alongside with the reported fair-to-moderate HR Rec relative reliability, discourage consideration of HR Rec as raw data difference values [23].
Cole et al. [3,8] provided longitudinal evidence of an association between decrements in HR Rec and reductions in all-cause and cardiovascular mortality. Furthermore, the same authors demonstrated the predictive strength of absolute HR cut-off values when considering HR 60 and HR 120 . Using the suggested cut-off values (i.e. �12 beats�min -1 ), we only found 3-6% of abnormalities in HR 60 across the field tests. The training intervention and the associated increase in VO 2max (~9%) and corresponding decrement in HR max (~3%) produced a remarkable reduction in the initial HR 60 abnormalities, when considering a highly demanding test like YYIE2 (Table 5). Indeed, no HR 60 abnormalities were detected in the participants when examining pre-intervention HR 60 after the YYIE1 and YYIR1 tests. Interestingly, only 3% of participants reported HR 60 abnormalities in YYIE1 HR 60 post-intervention, suggesting a sort of independence between cardiorespiratory fitness improvement and HR 60 abnormalities. The resulting occurrence of HR 60 abnormalities may have been the direct consequence of the participants' health-related inclusion criteria. However, the lower prevalence of abnormal HR 60 values at post-intervention might be considered as evidence of a possible positive effect of recreational football practice on reducing potential health-related risk factors. The reported test-related prevalence of HR 60 abnormalities has practical importance for preventive medicine that warrants future studies [3,8,10].
However, the normative value proposed by these authors was derived from a population of male inactive subjects who were approximately 20 years older than the recreational football participants considered in the present study, which promotes the interest in population-specific normative values for tracking abnormalities in HR Rec [3,10].
The intermittent nature of recreational football, involving high-intensity bouts of exercise interspersed with activities performed at lower intensity for recovery, or less demanding match-related actions, potentially would be effective for improving HR Rec [18]. However, this study results did not provide evidence for enhanced HR Rec as result of a typical recreational football intervention. This was probably the consequence of considering the training-induced variations in HR peak and HR max when profiling the HR Rec variables at post-intervention [40,47]. Further studies with larger samples and a mechanistic design, are warranted to understand the variation in heart physiology provided by a casually intermittent activity such as recreational football in different populations. Additionally, the use of a control group would be useful in future studies to fully understand the nature of HR Rec and related variables.
These findings question the use of HR Rec as indicator of cardiorespiratory fitness level (see if you agree) or training-induced changes in healthy subjects participating in a recreational football intervention or in recreational players during the training process. The observed large changes (26-43%) in field test performance suggest sensitivity in tracking the aimed enhancement of the individual aerobic performance in recreational football training interventions. Changes in YYIE1 and YYIR1 performance were moderately and significantly associated with changes in HR 60 (r = 0.36, P = 0.045) and HR 120 (r = 0.38, P = 0.034) , respectively. These results support the general picture of a limited association between changes in cardiorespiratory fitness level in recreational football players and HR Rec variables.
In the published literature, HR Rec assessment is reported as absolute values (beats�min -1 ) [3,7,13,23]. Despite the practical interest of considering absolute values, differences in individual HR max may provide biased data supposedly producing false positive results. Heart rate recovery normalisation using peak test HR or HR max may potentially be the solution for avoiding biased data that may affect training prescription and clinical diagnosis. However, in this study the deliberate use of absolute or HR max derived HR Rec did not provide differences in the information supposed to have clinical importance. Inter-subject variability in test HR peak (~6%) may have been the cause of the reported limited effect of data normalisation on the considered variables. The practical interest of the effect of reporting HR Rec data deserves further studies with populations of different age and cardiorespiratory fitness level.
Attention should be paid when considering cut-off values (i.e. 12 or 43 beats�min -1 ) to qualitatively characterise the risk of developing cardiovascular diseases [3,[8][9][10]. A bias-limiting indicator of the cardiorespiratory risk could be developed using individual maximal values. Unfortunately, the reduced data variability of this study's participants was not helpful in providing meaningful guidelines and further studies are warranted. However, the age range of this study's participants (5.6 years, standard deviation range 22 years) may have affected the results, as in age-independent groups subjects with higher HR max reported better absolute HR Rec . In this study, a moderate association was reported between HR max and HR Rec .
Future training studies should also investigate the possibility to find more meaningful HR Rec reference time points as suggested in a cross-sectional study by Ostojic et al [48]. This would be of specific interest, as short-term HR Rec (i.e., as short as 20s) has been reported to be faster in athletes of intermittent sports (i.e., basketball, soccer and team handball) with at least four years of participation in these sports [48]. Given that, knowledge about the applicability of the short-term HR Rec concept in previously untrained subjects participating in a training intervention, using exclusively intermittent exercise like recreational football, would be of great practical interest.

Conclusions
It was not possible with this study design to support the validity of tracking post-exhaustion HR Rec to estimate individual aerobic fitness (VO 2max ), either in the untrained or trained status (i.e. pre-and post-intervention, respectively). Indeed, no significant and practically important associations were found between HR Rec variables and recreational football players' VO 2max . This study's results are in line with those reported in the athletic populations, suggesting HR Rec as a "per se" physiological adaptation that is independent of VO 2max level and changes [13].
Given the interest of this issue for public health, further studies involving a larger number of participants followed for a longer time are warranted. From the practical point of view, HR Rec is a reliable variable in the short-term (i.e. 12 weeks), nevertheless, it is not associated with the improvement in aerobic fitness in this population of recreational football players.
Although this study's results refer to recreational football players, the information obtained may be of interest for all professionals dealing with health-enhancing strategies evaluated under field conditions. ΔHR Rec is considered as the variable for tracking the efficiency of the physiological processes that underpin HR Rec . This study's results suggest that caution is advised when considering ΔHR Rec , as this variable may be affected by the concomitant reductions in HR at exhaustion values and, consequently, bias the reported differences.