The aim of the current study was to establish the reproducibility of key parameters that correspond to PFO rates (i.e. VPFO and Fatpeak) during treadmill ergometry. The main results of this investigation were the high ICC and Pearson’s correlation coefficient computed for VPFO and Fatpeak, in addition to the correspondingly low CV (i.e. 0.98, 0.97, 5.0 %; and 0.90, 0.81, 7.0 % respectively). Moreover, the performed Bland-Altman analysis has revealed a small bias of −0.3 km/h between Fat-peak tests, with 95 % LoA for the reproducibility of VPFO of 0.9 km/h (i.e. -2 ± 8 % of VO2peak).
To our knowledge, the present investigation is the first to report on the reproducibility and day-to-day variability of both VPFO and Fatpeak during treadmill ergometry running. Hence, the current results reveal excellent values for the particular relative and absolute reliability indicators. The study group of Gmada  seems to be the first to have taken a more comprehensive statistical approach to assess the repeatability of Fatpeak. In their study, 12 sedentary, but otherwise healthy males performed a graded exercise test (5 stages of 6 min at 20, 30, 40, 50 and 60 % of the maximal aerobic power (MAP)) after a 12-h overnight fast. ICC and CV values for Fatpeak across test re-test trials separated by a time interval of 4 days were 0.97 and 5.0 %, respectively. The mean differences ± 95 % LoA for Fatpeak was 0.6 ± 7.2 W, indicating that 95 % of the intra-individual differences should be contained between −6.6 and +7.7 W. Based on these values, relative and absolute reliability of Fatpeak were deemed as highly reliable by the authors. Unfortunately, no further appraisal has been made to address the physiological plausibility or applicability of the given LoA. Three other investigations have employed similar submaximal graded protocols (i.e. similar stage increment and duration, plus the 12-h overnight food restriction prior to each bout), which were based either on the measured or on the theoretical MAP to establish the reproducibility of Fatpeak. Yet, conflicting findings have been reported. Pérez-Martin  reports a CV of 11.4 % for Fatpeak, and considered it satisfactory after assessing 10 overweight, but otherwise healthy male participants (no LoA analysis carried out). Similarly, Michallet  reports on CV values between 7 and 12 %. Here, the reproducibility of Fatpeak was assessed via two different gas exchange techniques in a group of 14 healthy and moderately trained participants (9 males, 5 females). More recently, Croci  assessed 15 healthy and moderately trained males, and computed CV values between 16 and 20 % for Fatpeak while implementing three different data analysis procedures. The authors additionally report a high intra-individual variability with mean differences ± 95 % LoA for Fatpeak (calculated with a P3 function) of −4 ± 32 % of VO2peak, indicating that 95 % of the intra-individual differences should be expected between −37 and +28 % of VO2peak. Two other investigations using different methodological approaches have addressed the reliability and/or variability of Fatpeak estimations. Achten  has advocated good reliability after assessing 10 healthy and moderately trained males as they performed an incremental test to exhaustion (test start: 95 W; stage increment and duration: 35 W every 3 min) on three different occasions and after a 12-h overnight fast. The CV for Fatpeak (% of VO2peak) was 9.6 %. The authors additionally report a root mean square error (typical error) and 95 % confidence interval for Fatpeak of 0.23 l/min (0.17 -0.34 l/min). Meyer  on the other hand, shows a large intra-individual variability for Fatpeak after assessing 21 healthy participants (10 males, 11 females) of varying endurance capacities. Nutrition was moderately controlled, but with no fasting required prior to the exercise bouts. The implemented incremental exercise protocol was nearly identical to the one currently used in our study (further appraisal on the protocol is given below). The mean differences ± 95 % LoA for Fatpeak was −13 ± 0.91 l/min (−3.9 ± 28 % of VO2peak). Hence, 95 % of intra-individual differences were to be expected between −1.04 and +0.78 l/min (−32 and +23 % of VO2peak). In this case, the large variability can be mostly attributed to the fact that only the end of each exercise stage was evaluated and not a continuous curve (i.e. whenever PFO switches from stage 2 to 3, for instance due to a small difference in the recorded rates, it will then result in a large difference in the equivalent % of VO2peak).
In the current study, the computed scores agree closely with those reported by Gmada , especially the CV, which has come noticeably lower then all of the other values reported in preceding analyses. As to the intra-individual (day-to-day) variability of Fatpeak, when expressed as % of VO2peak, our LoA values have been distinctly lower then those observed by Meyer  and Croci . However, whilst these results enable closer comparisons to some of those from other investigations, making reasonable inferences as to the physiological plausibility and practical applicability of these LoA has shown to be a challenging task. As implied by Croci , previous studies have deemed an intra-individual variability of ± 10 beats/min for HR at VPFO as acceptable, since this reflects a realistic margin in individuals who use HR for the monitoring of training intensity [7, 15]. Accordingly, in the present investigation this threshold has been sustained in most participants, with only three of them eventually exceeding the given cutoff (though by no more than 3 beats/min). Therefore, based on the strong aggregate of reliability indices and the generally lower intra-individual variability observed for the aforementioned physiological aspects (i.e. Fatpeak as % of VO2peak and HR at VPFO), we consider the present Fatpeak estimations as the most reliable and coherent to date. Furthermore, the employed treadmill running protocol may be used as a reliable tool to identify Fatpeak in moderately trained individuals, and according to the reported intra-individual variability values, serve as the basis for future investigational research.
In spite of that, its applicability for athletic training is still questionable. For instance, the high day-to-day variability for PFO (g/min) remains largely unexplained. In the current study, PFO recordings between Fat-peak tests differed by a minimum of 0.01 g/min (1 %) and a maximum of 0.28 g/min (45 %) among the participants, which is consistent with inter- and intra-individual patterns described in previous investigations [1, 15, 16]. On the grounds of this known variability for PFO, recent studies [25, 26] have questioned the practical applicability of prescribing exercise training based on Fatpeak, since it remains debatable whether prolonged exercise at Fatpeak can indeed be maintained with PFO rates. Therefore, it may be ultimately necessary for prospective studies (e.g. those looking at the sustainability of PFO during prolonged exercise bouts at Fatpeak) to consider the LoA (or simply the individual test re-test difference) for Fatpeak, VPFO and PFO. Then, based on that, delineate the ± intensities in which exercise bouts should be performed and eventually evaluate how this impacts the sustainability of PFO (i.e. also in accordance to the identified intra-individual variability of each person). Other questions in need of further research include: 1) What are the physiological determinants and additional intrinsic/extrinsic factors influencing the variability of fat oxidation rates during running, as well as in other types exercise? 2) How applicable, versatile and reliable is the current protocol across different cohorts of people (e.g. patients, untrained persons or professional athletes)?
To date, there have been a few investigations assessing the reproducibility of Fatpeak [6, 7, 13–16]. Though the majority of those have failed to make thorough statistical analyses by not providing indicators of both relative and absolute reliability for Fatpeak estimations (i.e. the degree to which individuals/variables maintain their position in a sample with repeated measurements; or the degree to which repeated measurements vary for individuals/variables), in addition to practical information on the respective intra-individual (day-to-day) variability by establishing the LoA (i.e. the individual subject differences in a test re-test plotted against the respective individual means) [18, 27–29]. Hereto, previous studies suggest that an ICC greater than 0.90 is reflective of high relative reliability, while values between 0.80 and 0.90 should be rated as moderate, with figures under 0.80 being graded as not sufficient for physiological testing [6, 30]. Additionally, a Pearson’s coefficient greater than 0.80 is advocated as high , whereas a CV under 10 % can be considered as an indicator for a reliable test, being a commonly used and accepted threshold for biological variables [6, 31, 32].
In the current study we have implemented rigid pre-testing conditions with standardized nutrition and exercise restraint for the 24 h prior to each submaximal bout. Yet, other methodological factors such as the elected exercise protocol, data analysis approach as well as the embedded equipment error may affect the determination of fat oxidation rates and subsequently VPFO . The currently employed exercise protocol intends to cover the realistic range for VPFO determination and takes into account important physiological aspects in its design to ensure gas exchange maintains steady state for as long as possible . The start velocity (VLT) corresponds to the first increase in blood lactate and can be considered as the upper border for the conduction of regenerative training. The end velocity (VRER) represents a metabolic state where energy supply is expected to yield solely from carbohydrate metabolism. Ultimately, three stages in between these metabolic markers should account for an accurate determination of VPFO [15, 21, 33, 34]. Additionally, we have chosen to create P3 curves, as it is a valid and widely used method that models the overall kinetics of fat oxidation for a more coherent representation of VPFO and PFO .
Here we would like to comment on the 11 participants that had their VPFO and Fatpeak computed during the warm up phase. One reason for this could of course be the rather moderate aerobic endurance capacity of participants, since in less trained individuals Fatpeak occurs at lower exercise intensities than in trained individuals . However, when looking at the individual raw fat oxidation rates, only 5 subjects have had indeed higher fat oxidation values during the warm up phase. The remaining 6, had their highest raw values recorded at the end of the first stage and were somewhat “drifted backwards” due to the applied P3 interpolation and how the curve-fit reacted upon the variables. Such a drift can also occur in the opposite way as depicted in Fig. 1c, which in this case, was caused when curve-fitting the overall means for fat oxidation rates instead of individual values. This prompted the curve into a small elongation (likely driven by the subjects that had PFO rates at the latter stages of the tests). Hence, the depiction of PFO rates that are slightly lower than the mean of individually interpolated values, and which also occur during the test phase and not the warm up. Still, the use of a mathematical model such as the P3, is a more consistent approach than just accounting for the raw measured values when analyzing data that does not align in a perfect curve . However, alternative ways of curve-fitting might be evaluated in the future.
At last, it must be noted that the total variation observed in our test re-test is a sum of both biological and equipment variation (error) [15, 16]. Though analyzing the relative contribution of each of these parameters was beyond the scope of this study, the used gas exchange analyzer has been considered reliable . Ideal ICC values (1.00) were computed for ventilation (VE) VO2 and VCO2. Respectively, the average intra-device technical error of measurement (%TEM) was 0.2, 1.4 and 1.1 %.