Following data collection, correction, and detrending, the vast majority of HRV research employs time-domain and/or frequency-domain HRV measures. Regarding the time-domain (computationally the simplest and most consistent analysis method across studies), the two most common measures are the standard deviation of R-R intervals (SDRR), a measure of overall variability; and the root mean square of successive differences of R-R intervals (RMSSD), a measure of beat-to-beat variability. The latter is sometimes calculated slightly differently as the standard deviation of successive differences (SDsd).

Frequency-domain measures express HRV as a function of frequency, rather than time, since different spectral power components of HRV might relate to different elements of cardiac autonomic activity (Akselrod et al., 1985). There are different methods (and multiple sub-variations of methods) utilized to calculate HRV spectra, with the two most common employing a Fourier transform or autoregressive modeling. More recently, frequency-domain analysis has been applied to non-stationary signals using time-varying methods such as short-time Fourier transform (STFT) analysis (Cottin et al., 2006; Martinmaki et al., 2006; Kaikkonen et al., 2010, 2012). Regardless of the method used, the primary components are low frequency (LF, often 0.04–0.15 Hz) and high frequency (HF, often 0.15–0.40 Hz) spectra. Very low frequency spectra (VLF, <0.04 Hz) is seldom reported. Together these constitute total power (TP). These may be expressed as absolute power (ms²) or as power spectral density (ms².Hz⁻¹). Additionally, the ratio of LF to HF power (LF:HF) is also often reported, and HF and LF may also be normalized to TP (which may or may not include the VLF component) giving HF-nu and LF-nu (nu = normalized units). Variations of the upper and lower limits of each frequency band have also been employed (Radaelli et al., 1996; Avery et al., 2001; Pichon et al., 2004; Povea et al., 2005; Casties et al., 2006; Spadacini et al., 2006; Martinmaki et al., 2008).

Finally, HRV measures are often observed to be non-normally distributed. For this reason, a data transformation (typically natural-logarithm, Ln) is sometimes applied to yield an approximately normal distribution and permit parametric statistical analysis.

In addition to time and frequency domain HRV, other approaches of expressing the variability of HR are sometimes employed. Poincaré plots warrant mention because the measure “standard deviation 1” (SD1) is used to quantify short-term beat-to-beat HRV (similar to RMSSD). In fact, SD1 is a scaled derivative of SDsd (which is very similar to RMSSD), i.e., SD1 = SDsd ÷ √2 (Brennan et al., 2001). Indeed, in studies reporting SD1 and SDsd or RMSSD, manually dividing the reported SD1 results by √2 gives nearly identical results to those reported for SDsd (Tulppo et al., 1996) and RMSSD (Leicht et al., 2008). Other measures of HRV based on non-linear dynamics are less often utilized, such as power-law analysis, entropy, dimensionality and fractals (Oida et al., 1997; Huikuri et al., 2003, 2009; Perkiomaki et al., 2005; Baumert et al., 2009). Some reports have suggested that non-linear HRV measures may provide independent risk stratification and prognostic insight to complement time and frequency domain HRV measures (Makikallio et al., 1999; Huikuri et al., 2003; Stein et al., 2005), as well as offer certain advantages such as not requiring a stationary signal. In particular, fractal scaling exponents derived from detrended fluctuation analysis (DFA) might be of clinical value (Peng et al., 1995). However, Francis et al. (2002) suggested that fractal scaling exponents are fundamentally related to a weighted form of frequency-domain analysis. Regardless, the short term scaling exponent (α1) derived from DFA during incremental exercise may be modified by training (Karavirta et al., 2009). Moreover, it is interesting that the “inverted-U” bi-phasic response of DFAα1 during incremental exercise (Hautala et al., 2003) appears to be generally consistent with the bi-phasic nature of parasympathetic reflex control of HR as a function of exercise intensity (White and Raven, 2014). Notwithstanding the potential value of non-linear HRV measures (which warrant further investigation), these techniques are beyond the scope of this review.

Interpretation of HRV as reflecting certain aspects of cardiac autonomic activity is complicated by the fact that rather than being a direct measure of autonomic nerve activity, HRV quantifies the modulation of the end-organ response, i.e., HR. HR is under the dual control of direct sympathetic and parasympathetic innervation at the sino-atrial (SA) node, but is also influenced by sympatho-adrenal activation and non-autonomic effects such as mechanical/hemodynamic influences and local reflexes (Bernardi et al., 1990; Rowell et al., 1996). Furthermore, complex non-linear pre-junctional and post-junctional parasympathetic-sympathetic interactions have been reported (Levy, 1971; Kawada et al., 1996; Sunagawa et al., 1998; Uijtdehaage and Thayer, 2000; Miyamoto et al., 2003, 2004; Paton et al., 2005), whereby activity of one branch can in some instances augment and in other instances attenuate the activity and/or effect of the other branch. It is therefore critical to appreciate that HRV is an indirect indicator of cardiac autonomic modulation.

For clinicians and exercise/sport scientists, the primary interest in HRV relates to (a) its prognostic potential value in cardiac disease and sudden cardiac death (Camm et al., 1996; Tsuji et al., 1996; Kikuya et al., 2000), and (b) the general acceptance of beat-to-beat measures (cPNA-HRV) as indicators of cardiac parasympathetic modulation. Indeed, notwithstanding some interpretative controversies, these measures are regularly employed in this manner. The underlying mechanisms and interpretation of other HRV measures (such as LF, LF-nu, and LF:HF) are less established, likely resulting from complex sympathetic-parasympathetic interactions. While this has not prevented the use of these measures as indices of sympathetic activity or “sympatho-vagal balance,” the majority of evidence does not support this approach. In light of interpretative controversies, cPNA-HRV measures perhaps provide qualitative or ordinal (rather than parametric quantitative) insight into cardiac parasympathetic activity (Billman, 2006).

Several studies have investigated the effect of exercise intensity on HRV during exercise. In addition to different modalities being utilized between these studies, the duration for which participants exercised at each intensity varies greatly, such as 2 min (Tulppo et al., 1996; Martinmaki et al., 2008), 3 min (Tulppo et al., 1998, 1999; Hautala et al., 2003; Karapetian et al., 2012), 5 min (Radaelli et al., 1996), 8 min (Casties et al., 2006; Michael et al., 2016), 10 min (Martinmaki and Rusko, 2008) and 15 or more min (Yamamoto et al., 1991; Saito and Nakamura, 1995; Leicht et al., 2008; Boettger et al., 2010). Nevertheless, an analysis of the literature allows us to identify some general responses for some HRV measures as a function of exercise intensity.

Compared to the effect of exercise intensity, few studies have investigated the effect of exercise duration HRV responses during exercise. This may be in part due to the pronounced effect of intensity on HRV, which has two important implications for investigating other influences on HRV during exercise: (1) Due to the HRV-intensity relationship (whereby most HRV measures reach a near-zero minimum at/above moderate intensity), any potential influence of duration is minimized; and (2) During prolonged exercise HR may change despite no change in external load, i.e., cardiovascular drift (Montain and Coyle, 1992), meaning that “internal intensity” is important to consider. Thermoregulatory factors (in particular fluid losses from sweating) play a key role in cardiovascular drift as exercise is prolonged (Coyle and Gonzalez-Alonso, 2001), with significant cardiovascular autonomic implications due to changes in blood pressure and blood flow redistribution (Figure 1).

Kaikkonen et al. (2007) investigated the effect of exercise duration on frequency domain HRV measures during different exercise intensities in sedentary women, who performed low intensity (~45% VO₂max) and high intensity (~77% VO₂max) exercise for two different durations (approximately 38 vs. 76 min for low intensity and 30 vs. 60 min for high intensity). Despite a doubling of exercise duration, no significant difference in HRV during exercise was observed for either intensity. Similar findings were reported by the same authors in a later study (Kaikkonen et al., 2010) investigating the effect of a 300–400% increase in exercise duration, in which recreational-level athletes ran at ~66% VO₂max for ~20 or ~90 min.

Pichon et al. (2004) investigated frequency domain HRV measures during different short-duration exercise bouts, i.e., 3, 6, and 9 min of exercise at 60 and 70% of power reached at VO₂max (PVO₂max), and 3 and 6 min at 80% PVO₂max. HF was higher during 3 min compared to 6 and 9 min. LF also decreased with increased exercise duration. No significant effects were observed for normalized measures or LF:HF. However, as all three intensities were of at least moderate-high intensity, the heart rates differed substantially between exercise durations. The lowest intensity (60% VO₂max) elicited a mean HR of 143, 161, and 167 b.min⁻¹ at 3, 6, and 9 min, respectively. These differences were enhanced with higher intensities, likely in part because of exercise onset HR kinetics and the time required to reach steady state. Thus, despite these bouts being matched for relative intensity (in terms of %PVO₂max), the elevated HR associated with higher intensity exercise of such short durations means it is difficult to separate the independent effect of exercise duration from that of exercise intensity in this study. Similarly, Moreno et al. (2013) investigated time and frequency domain HRV throughout 90 min of moderate intensity exercise (60% VO₂peak). Despite HRV measures being greatly reduced by the time of the initial exercise recording (25 min), the authors found that HR increased throughout 90 min of exercise (from ~140 to ~150 b.min⁻¹), while SDRR, RMSSD, LF and HF were further decreased by 90 min.

Relatively few studies have compared the HRV responses during different modalities of dynamic exercise. In addition to the methodological challenge involving the strong HRV-intensity relationship, it is difficult to standardize exercise “intensity” when comparing different modalities. Some of these issues relate to what basic metric to employ (e.g., power output, HR, or VO₂). For example, arm crank exercise will elicit a higher HR when compared with leg cycling for the same rate of energy expenditure (same absolute VO₂; Tulppo et al., 1999) or work rate (Leicht et al., 2008). There is also the issue of whether to match relative or absolute intensity (and whether this is mode-specific). For instance, peak HR will likely be 10–20 b.min⁻¹ lower during arm cranking compared with leg cycling (Tulppo et al., 1999; Ranadive et al., 2011). The physiological response (e.g., HR or VO₂) elicited for any intensity of arm crank exercise is likely not equal to response elicited during leg cycling exercise, despite the workload representing the same percentage of mode-specific maximum HR or maximum VO₂. Because of these inherent methodological challenges, the effect of modality on HRV during exercise has been investigated with varying approaches.

A small number of studies have investigated the test-retest reliability of HRV measures during exercise. Apart from aforementioned disparities amongst the literature regarding exercise protocols and HRV analysis methodology, there are also differences in the assessment of reliability. The intraclass correlation coefficient (ICC) reflects the ability of a test to differentiate between individuals and is therefore considered a measure of relative reliability, although there are multiple methods of calculating ICC (Weir, 2005). Alternatively, measures such as the coefficient of variation (CV), typical error of measurement (TEM, which may be expressed as a CV), and Bland-Altman limits of agreement (LoA) may be considered measures of absolute reliability as they reflect the trial-to-trial noise. Carrasco et al. (2003) demonstrated that the ICC for a range of time and frequency domain measures (Ln-transformed) during cycling exercise at 80 W ranged between 0.70 and 0.91, which was comparable to resting measures. Similar ICC-values (~0.9) were reported for Ln-transformed time domain measures during slow walking at 4 km/hr in another study (Boullosa et al., 2014). These authors also reported that that the TEM(CV) was 16–22%. Utilizing a Bland-Altman approach, Tulppo et al. (1998) demonstrated that the LoA were lower during cycling exercise for time and frequency domain measures compared to rest, although this may be partly due to the substantially reduced values during exercise compared with rest.

During exercise, HRV measures demonstrate a curvilinear decay as a function of exercise intensity, that is closely related to exercising HR. HRV measures associated with cardiac parasympathetic activity (e.g., RMSSD and HF) usually reach a near-zero minimum at moderate intensity (possibly being associated with the first ventilation/lactate threshold). These measures are sometimes observed to increase slightly as exercise intensity increases toward maximum, although this is likely mediated by non-neural mechanisms such as direct mechanical effects of respiration on the SA node. The data also leads to further questioning of the use of frequency domain ratio and normalized measures as indicators of sympathetic activity or “sympatho-vagal balance.” In addition to demonstrating inconsistent responses to exercise, the response of these measures is rarely consistent with our current understanding of autonomic control during exercise, namely progressive parasympathetic withdrawal and sympathetic activation (White and Raven, 2014).

Regarding exercise duration, the limited body of literature suggests that prolonged exercise duration can influence (attenuate) HRV during exercise, although this has only been observed in studies where there is a concomitant increase in HR (i.e., cardiovascular drift) and also when HRV has not already reached the intensity-dependent minimum. Indeed, this elevated HR (representing elevated internal intensity, regardless of the external load) may be the cause for any change in HRV, rather than any direct effect of exercise duration. Conversely, prolonged exercise duration may be associated with progressive parasympathetic withdrawal (indicated by lower HRV), and this in turn might contribute to cardiovascular drift (Kukielka et al., 2006). Regardless of the direction of any cause-effect relationship, it is difficult to infer what affects HRV under conditions of different HR-values, as HR likely has a purely mathematical effect on HRV (Billman, 2013a; Sacha, 2013, 2014), whereby a greater HR can reduced HRV despite no change in the actual variability of autonomic outflow.

While there is limited data available, the evidence suggests that exercise modality can modify HRV, although not all research supports this. Interpretation of results is difficult as any influence of exercise modality is confounded by the issue of matching for exercise “intensity.” Interestingly, studies investigating different dynamic modalities that would be expected to elicit different orthostatic/circulatory conditions (e.g., posture, gravity, or water immersion) have typically found no substantial effect on HRV during exercise. Instead, the muscle group employed and the mode of contraction (e.g., static vs. dynamic) seem to be stronger modality-related factors influencing HRV response during exercise. Nevertheless, exercise intensity is the strongest determinant of HRV during exercise.

Several studies have investigated HRV during recovery from different exercise intensities. A recent review (Stanley et al., 2013) quantitatively summarized the findings of some of these studies demonstrating that a higher exercise intensity is associated with a slower recovery of cPNA-HRV measures, specifically Ln-transformed RMSSD or HF (Furlan et al., 1993; Terziotti et al., 2001; Mourot et al., 2004; Parekh and Lee, 2005; Niewiadomski et al., 2007; Seiler et al., 2007; Kaikkonen et al., 2008). These data are redrawn in Figure 4. Other studies have reported overall similar findings for cPNA-HRV and other HRV measures, namely that a greater exercise intensity results in a slower HR and HRV recovery (Perini et al., 1990; Hayashi et al., 1992; Buchheit et al., 2007a; Kaikkonen et al., 2007, 2010, 2012; Martinmaki and Rusko, 2008; Gladwell et al., 2010; Al Haddad et al., 2011; Casonatto et al., 2011; Dupuy et al., 2012). From a mechanistic standpoint, the effect of intensity on HRV recovery is likely associated with the amount of non-oxidative energy contribution and subsequent stimulation of the muscle metaboreflex (Buchheit et al., 2007a).

However, the details of the intensity dose-response has not been clearly elucidated, particularly during the initial 10 min following exercise when comparing moderate vs. high exercise intensities. Most studies investigating HRV following more than two intensities report differences between some, but not all exercise intensities (Kaikkonen et al., 2007; Seiler et al., 2007; Gladwell et al., 2010; Casonatto et al., 2011). Seiler et al. (2007) suggested that the first ventilation threshold might demarcate an autonomic “binary threshold” regarding HRV recovery, whereby exercise below this intensity is associated with rapid HRV recovery while exercise above this intensity results in delayed HRV recovery which is intensity-independent (at least in highly trained athletes). In contrast, inspection of Figure 4 suggests that exercise intensity may elicit a more graded (rather than binary) effect on HRV recovery (at least during the first hour post-exercise). This latter interpretation is consistent with the findings of a recent study by this research group demonstrating a graded response following three different exercise intensities (Michael et al., 2016).

Similar to responses during acute exercise, few studies have investigated the effect of exercise duration on HRV during recovery after exercise, in particular when controlling for the intensity. Seiler et al. (2007) investigated the recovery of HRV following either 60 or 120 min of low intensity running exercise (below VT1 at ~60% VO₂max) in a highly trained athletic population. By the earliest time-point investigated post-exercise (5–10 min), HRV (including RMSSD) had recovered to pre-exercise levels and there was no significant effect of duration on HRV measures throughout the 4 h recovery period. These findings are consistent with those reported by Casonatto et al. (2011), who investigated the effect of exercise duration on HRV during 60 min of recovery by having healthy participants cycle for 30 min or ~45 min at 60% VO₂max. There was no significant duration effect for the HRV measures assessed during recovery (RMSSD, LF-nu, HF-nu, and LF:HF).

The findings of Kaikkonen et al. (2007) also suggest a lack of any duration effect on HRV during the immediate recovery period. In that study, the time-course of immediate HRV recovery in sedentary women was investigated using STFT frequency-domain analysis. The participants performed low intensity (~45% VO₂max) and moderate-high intensity (~77% VO₂max) running for ~38 vs. ~76 min and ~30 vs. ~60 min for low and moderate-high intensity, respectively. Despite a strong effect of exercise intensity on the time-course of HRV recovery from exercise, there was no significant effect of exercise duration for either intensity from 1 to 30 min recovery.

While these studies suggest that a 100% increase in exercise duration does not alter HRV recovery following exercise, the results of another study by Kaikkonen et al. (2010) indicate that exercise duration may influence post-exercise HRV when duration is extended by 300–400%. In that study, recreation-level athletes ran at ~66% VO₂max for ~20 and ~90 min. During the initial 3 min of a 15 min recovery, the longer duration resulted in lower Ln-LF, Ln-HF, and Ln-TP.

Very few studies have examined post-exercise HRV recovery following different modalities of “aerobic” exercise. Cunha et al. (2015) investigated the immediate recovery period following incremental exercise of three modalities: walking, cycling, and running. During the 5 min recovery period, it was reported that HR recovery as well as the recovery of RMSSD was more rapid following exercise involving a smaller muscle mass or energy expenditure (i.e., cycling > walking > running), with the authors therefore concluding that muscle mass and/or energy expenditure are determinants of post-exercise parasympathetic reactivation. Similarly, active children demonstrated more rapid HR and HRV recovery following maximal incremental arm-cranking vs. cycling (Ahmadian et al., 2015), The findings of these two studies were also consistent with those of other studies investigating HR recovery (without HRV measurement) following exercise mode comparisons such as cycling vs. running (Rahimi et al., 2006; Maeder et al., 2009) and cycling vs. arm-cranking (Ranadive et al., 2011). These studies also utilized maximal incremental exercise. To our knowledge, HRV recovery has not been investigated following different exercise modalities utilizing submaximal intensity.

During acute post-exercise recovery, the few studies investigating test-retest reliability have reporting conflicting results. Some studies report moderate to good relative reliability, such as ICC-values of 0.58–0.91 during active recovery (Boullosa et al., 2014) and 0.69–0.92 during static recovery (Carrasco et al., 2003) for various HRV measures. Alternatively, Dupuy et al. (2012) reported ICC-values ranging from 0.14 to 0.97 for during static recovery. Absolute reliability has also been reported to vary markedly, with TEM(CV) reported as 19–27% (Boullosa et al., 2014) and 8–65% (Al Haddad et al., 2011), while Dupuy et al. (2012) reported CVs ranging from of 27 to 141%. Considering the methodological differences amongst these few studies, it is difficult to draw any conclusions regarding the test-retest reliability of HRV during recovery.

A large body of research indicates that, as is the case during exercise, exercise intensity is a primary determinant of the immediate post-exercise recovery of HRV. Upon exercise cessation, HRV measures demonstrate a time-dependent recovery that is usually (though not always) slowed following a greater preceding exercise intensity. Thus, the intensity-dose response on HRV during recovery has not been clearly elucidated.

It is also not yet clear how exercise duration affects HRV during post-exercise recovery. While three studies report that a 100% increase in exercise duration does not alter HRV during immediate recovery, one study found that HRV recovery was slowed following a 300–400% increase in exercise duration (from ~20 to ~90 min). This may suggest that exercise duration must be prolonged beyond some critical length (either relative or absolute) before an effect on HRV recovery might be observed, however this remains speculative. Additionally, it is not clear how preceding exercise intensity and duration may interact to influence post-exercise HRV recovery. Finally, the two studies identified that investigated the effect of exercise modality on post-exercise HRV suggest that a greater active muscle mass and/or energy expenditure is associated with a slower HRV recovery, at least following maximal incremental exercise. Clearly, additional studies are needed to further elucidate the influences of exercise duration and modality on post-exercise HRV responses.

STI outcomes—such as the pre-ejection period (PEP), left ventricular ejection time (LVET), and PEP-to-LVET ratio (PEP:LVET)—are measures of cardiac performance that may be non-invasively measured using techniques such as bio-impedance cardiography. In particular, PEP may be used to assess cardiac sympathetic β-adrenergic activity (Ahmed et al., 1972; Sherwood et al., 1990; Cacioppo et al., 1994). The physiological rationale derives from sympathetic activity eliciting positive inotropic effects on the ventricular myocardium, thus increasing cardiac contractility and resulting in a more rapid development of force and intraventricular pressure. This decreases the time required to reach aortic pressure (the isovolumetric contraction time, IVCT) and therefore opening of the aortic valve, thereby attenuating PEP (hence PEP is inversely associated with cardiac sympathetic activity). In contrast to being richly innervated by cardiac sympathetic neurons, the ventricles are not extensively innervated by parasympathetic neurons, meaning that changes in contractility indices such as PEP are largely attributed to changes in cardiac sympathetic activity. Indeed, sympathetic stimulation by β-adrenergic agonists shortens PEP (Harris et al., 1967; Ahmed et al., 1972; Schachinger et al., 2001), whereas this response is reduced or abolished under conditions of β-adrenergic blockade (Harris et al., 1967; Benschop et al., 1994; Cacioppo et al., 1994). PEP:LVET has also been utilized as a measure of cardiac contractility, however LVET and PEP:LVET are influenced by the underlying HR (Cokkinos et al., 1976; Cacioppo et al., 1994) as well as parasympathetic activity (Cacioppo et al., 1994). In contrast, PEP is not substantially altered by changes in HR (Harris et al., 1967; Cokkinos et al., 1976). Thus, while it is widely acknowledged that LVET and PEP:LVET should be corrected for the underlying HR, the need for HR-correction of PEP (Weissler et al., 1968; Lewis et al., 1982) is contentious (Cokkinos et al., 1976; Spodick et al., 1984; Rousson et al., 1987; Cacioppo et al., 1994). As a result, PEP appears to be the STI measure of choice for reflecting cardiac sympathetic β-adrenergic activity (Rousson et al., 1987; Cacioppo et al., 1994).

The interpretation of PEP as reflecting cardiac sympathetic activity is not without limitations. In particular, non-sympathetic influences which may alter PEP need to be appreciated. Although the parasympathetic influence on the ventricles is considered to be weak, it is possible that parasympathetic outflow may inhibit positive inotropic sympathetic effects when sympathetic outflow is high (Levy, 1971; Azevedo and Parker, 1999; Figure 1). Another confounding factor relates to the fact that PEP encompasses the electromechanical delay as well as the IVCT. However, ~75% of the shortening of PEP in response to β-adrenergic agonists is due to shorting of IVCT (Harris et al., 1967), while changes during exercise and recovery are almost entirely accounted for by IVCT (Nandi and Spodick, 1977). In addition to these factors, cardiac loading may elicit non-sympathetic influences in PEP, in particular during exercise and post-exercise recovery. Namely, the Frank-Starling mechanism (length-dependent increase in cardiac contractile force, i.e., preload) and/or a decrease in aortic pressure (afterload) may shorten PEP (Lewis et al., 1977, 1982; Buch et al., 1980; Joubert and Belz, 1987). However, the extent to which preload and afterload actually change during exercise and recovery is debated (Plotnick et al., 1986; Kimball et al., 1993; Warburton et al., 2002; Rowland, 2008; La Gerche and Gewillig, 2010), as is the extent to which PEP reflecting cardiac sympathetic influences may be confounded (Nandi and Spodick, 1977; Cousineau et al., 1978; Lewis et al., 1982; Obrist et al., 1987). Future research is needed to clarify the extent to which these potential confounding factors influence PEP during exercise and recovery.

There are also methodological issues to consider. For example, bio-impedance cardiography is highly sensitive to movement artifact, which can make reliable signal acquisition difficult during exercise involving substantial upper body movement. Furthermore, postural changes as well as large and rapid thoracic movements (e.g., heavy breathing during hard exercise) can also influence thoracic impedance and make waveform identification and interpretation difficult.

The response of PEP to exercise is generally consistent with what would be expected of a true (inverse) cardiac sympathetic indicator, i.e., an intensity-dependent decrease from rest to exercise (Ahmed et al., 1972; Nandi and Spodick, 1977; Miyamoto et al., 1983; Miles et al., 1984; Smith et al., 1989). PEP has rarely been investigated during post-exercise recovery. Nandi and Spodick (1977) demonstrated that the gradual recovery of PEP during the initial 5 min post-exercise period is intensity dependent across a range of absolute submaximal intensities (50, 100, and 150 W), i.e., greater intensity elicited a slower recovery. PEP:LVET has also been reported to be lower following maximal (compared with submaximal) exercise (Crisafulli et al., 2006b), but not following different submaximal exercise intensities (Nandi and Spodick, 1977; Crisafulli et al., 2006b). Regardless, the response of PEP during exercise is consistent with our understanding of intensity-dependent sympathetic activation. Very limited data suggests that post-exercise PEP recovery is also consistent with intensity-dependent sympathetic withdrawal.

No study has investigated the effect of exercise duration on STI outcomes either during exercise or post-exercise recovery. Regarding the effect of exercise modality, inspection of the data reported by Miles et al. (1984) suggests that PEP may be lower during arm-crank exercise compared with cycling for a HR range of ~80 to ~150 b.min⁻¹, however this is not clear due to different exercising heart rates. Alterations in posture (e.g., supine vs. upright cycling) would be expected to alter cardiac loading (in particular preload). Consistent with this, PEP during supine (compared with upright) cycling was reported to be lower at the same absolute intensities of 50 and 100 W, despite a lower HR at 100 W (Miyamoto et al., 1983), although a lack of any significant effect of posture on PEP during exercise has also been reported (Smith et al., 1989). Additionally, Crisafulli et al. (2006a, 2008) demonstrated that post-exercise muscle occlusion elicits an intensity-dependent delay in STI recovery. As for cPNA-HRV measures of post-exercise parasympathetic reactivation (see previous section), these findings highlight the important role that the muscle metaboreflex likely plays in regulating the post-exercise recovery of STI measures reflecting sympathetic withdrawal.

STI measures (in particular PEP) are a useful non-invasive indicator of cardiac sympathetic activity. As is the case for HRV, STI measures reflect the integrated end-organ response, in this case indirectly assessing inotropic cardiac effects that are understood to be under strong sympathetic influence. Accordingly, interpretative caveats/limitations relating to non-sympathetic influences (particularly potential cardiac loading effects, i.e., preload and afterload) need to be appreciated. Notwithstanding these limitations, the measurement of PEP likely provides valuable non-invasive insights into cardiac autonomic regulation, as PEP is strongly (inversely) associated with cardiac sympathetic activity.

The effect of different exercise dosages on the response of PEP (or other STI outcomes) during exercise has not been extensively investigated. Nevertheless, available data indicates that the (inverse) response of PEP is consistent with our understanding of sympathetic activity during exercise and recovery. Additional studies are needed to confirm this, particularly during the recovery period. It is not known how exercise duration influences PEP responses to exercise and recovery, nor is it well established how these responses are altered by different exercise modalities. Additionally, the test-retest reliability of these measures during and following exercise requires further research.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.