Twenty-four able-bodied upper-body-trained male individuals (age 28.3 ± 9.3, body mass 77.4 ± 8.9 kg, and height 1.8 ± 0.1 m) participated in this study. Participants were mainly cross-country skiers (N = 23) and additionally one rower who regularly trained cross-country skiing, all of whom participated in recreational or national level cross-country skiing and rowing races, respectively. All were highly trained with a running VO_2max of 66 ± 7 mL·kg⁻¹·min⁻¹ (range 53.0–75.9) and an average of 39 ± 11 (range 22.5–75) training hours per month (based on self-reported training hours from their training diary logs; www.olt-dagbok.net), most of which was endurance training and a considerable part employing the upper-body. The participants were instructed to refrain from heavy training and alcohol consumption 24 h before the start of the testing, caffeine intake the day of the testing and food intake 2 h before. A questionnaire was filled out on each day to monitor if the participants followed these instructions, as well as to exclude any prior illness or injury that might have interfered with the testing. The study was pre-approved by the Regional Committee for Medical and Health Research Ethics of Mid-Norway and conducted in accordance with the Declaration of Helsinki. All participants signed an informed consent form prior to participation in the experiment and were made aware that they could withdraw from the study at any point without providing an explanation.

The testing consisted of two test days, where participants performed a 1-min and a 3-min all out and an incremental test to exhaustion in an upper-body poling mode on a Concept2 ski-ergometer (Concept2, Inc., Morrisville, USA). Each participant performed the tests in the same order and at the same time of the day (to minimize the bias of diurnal variation in performance; Atkinson and Reilly, 1996). The test order was randomized between participants. The time between test days was a minimum of 48 h and an average of 4 ± 3 days (range 2–11 days). Before the start of the testing on the first day, the participants' body mass was assessed by the built-in weighing scale of a bioelectrical impedance analyzer (Inbody Co., Ltd., Seoul, Korea).

After being equipped with an oro-nasal mask (Hans Rudolph Inc, Kansas City, MO, USA) and a heart rate (HR) monitor (Polar Electro Inc., Port Washington, NY, USA), the participants tightly strapped themselves around the hips and thighs into a seat construction (see Figure 1). They were then familiarized with the test setup and mode by performing four times 5-min submaximal stages mode at an overall rate of perceived exertion of 9 (very light), 11 (light), 13 (somewhat hard), and 15 (hard) on a 6–20 Borg scale (Borg, 1982; Shephard et al., 1992).

A 15-min break followed the submaximal familiarizing stages, before a standardized 5-min warm-up preceded each of the peak tests, consisting of three and two min at the power output of the third (RPE 13) and fourth submaximal stage (RPE 15), respectively. The third minute of warm-up was inter-dispersed by two 5-s sprints at 90–95% of maximal sprint power. Both, the 1-min and the 3-min test were self-paced closed-end tests with the instruction to find a power output that the participants thought they could maintain throughout the test. Three pacing strategies were possible: (1) a higher power output at the start with a drop toward the end (positive pacing), (2) a stable power output throughout the test (even pacing), and (3) a lower power output at the start with an increase toward the end (negative pacing) (Atkinson et al., 2007). Positive and negative pacing were defined as more than a 10% increase or decrease of the last 30-s average as compared to the initial 30-s average power output. The incremental test started at the individual power output of the third submaximal stage (rounded to the nearest 5-W value) and participants were instructed to continuously increase power output by 10 W every 30 s. Between each of the maximal tests, a rest period of 26 ± 3 min was given and the participants were optionally allowed to drink water or sports drink. It has previously been shown that a recovery period of 20 min between maximal tests allows participants to maintain performance (Weltman et al., 1979; Vesterinen et al., 2009; Moxnes and Moxnes, 2014).

VO₂, VCO₂, and VE were measured breath-by-breath using a spiroergometer (Oxycon Pro, Jaeger, Viasys BV, CA, USA) which was calibrated against a known mixture of gases (5% CO₂, 15% O₂) and a known air flow (from a 3 L syringe) prior to each test. HR was continuously recorded during the tests. A blood sample was taken 1 and 3 min after each test and blood lactate analyzed by a Biosen C-Line Sport lactate measurement system (EKF-diagnostic GmbH, Magdeburg, Germany). Overall (RPE_O), respiratory (RPE_R) and muscular rate of perceived exertion (RPE_M), and were recorded after each test as described more in detail by Shephard et al. (1992). Power output per stroke was recorded by the skiergometer's internal software (Concept2, Morrisville, USA) and recorded by a Sony Alpha 58 video camera (Sony Corporation, NY, USA).

A minimum number 21 participants was determined by a-priori analyses in G^*Power 3.1, with an effect size of 0.65 (calculated as a Cohen's d based on VO_2peak values in a similar sample from Baumgart and Sandbakk, 2016, an alpha level of 0.05 and a power of 0.80. Two of the 24 participants were not able to complete the 1-min and the 3-min test on the second day. In these two tests, the data from 22 participants were analyzed. Breath-by-breath respiratory data was interpolated at individually fitted sample frequencies, resampled at 1-s intervals and 10, 30, and 60-s averages were calculated in MATLAB 8.1.0. (R2016a; Mathworks Inc., Natick, MA) The single highest 30-s average value was then identified as VO_2peak as recommended by Robergs et al. (2010) and used in the further analyses. The highest 10 and 60-s averages were used to investigate if changes in averaging procedure affected the results. Moving 3-s averages were calculated for the HR data and the highest value defined as peak heart rate (HR_peak). The higher of the two blood lactate values was defined as peak blood lactate (BLa_peak). Thirty seconds averages were calculated for the PO data and the highest value defined as peak power output (PO_peak).

Data are presented as mean ± SD unless specified differently and an α level of 0.05 was employed to indicate statistical significance. All calculations and statistical tests were executed in Microsoft Excel (Version 2010, Microsoft Cooperation, The Microsoft Network, LLC, Richmond, USA) or in SPSS 22.0 (Software for Windows, SPSS Inc., Chicago, IL, USA).

Individual differences and mean values of day 1 to day 2 and corresponding limits of agreement are visualized in Bland-Altman plots in Figure 2 and displayed in Table 1. All data used in the analyses of this study are found in Datasheet 1. Based on the average values of test day 1 and 2, the incremental (45.4 ± 5.5 mL·kg⁻¹·min⁻¹, 196 ± 28 W) and the 3-min test (44.5 ± 5.5 mL·kg⁻¹·min⁻¹, 201 ± 36 W) resulted in significantly higher VO_2peak and lower PO_peak as compared to the 1-min test (40.4 ± 5.0 mL·kg⁻¹·min⁻¹, 256 ± 47 W) (all p < 0.001). Additionally, the incremental test resulted in significantly higher VO_2peak (p < 0.03) (see Supplementary Figure 1). A plateau in VO_2peak (2 consecutive 30-s values within 2 mL·kg⁻¹·min⁻¹) was observed in ~80% of tests both during the 3-min and the incremental protocols of day 1 and 2, without any difference between test protocol or order of test day.

As compared to the 30-s average used in the above, employing a 10 or 60-s average would have resulted in significantly higher or lower VO_2peak, respectively, in all three tests (all comparisons p < 0.001). When using 10-s averages, the difference in VO_2peak between the tests would have remained unchanged (1-min: 41.3 ± 5.3, 3-min: 45.4 ± 5.5, incremental: 46.3 ± 5.6 mL·kg⁻¹·min⁻¹) compared to using 30-s averages. However, using 60-s averages, the VO_2peak difference between the 1-min as compared to the 3-min and the incremental test would have increased, and the differences between the 3-min and the incremental test decreased (1-min: 32.6 ± 4.2, 3-min: 43.8 ± 5.7, incremental: 44.2 ± 5.4).

High and very high ICCs across all physiological outcome parameters and PO_peak were found for the 1-min, the 3-min and the incremental test (Table 2). In all three tests, the SDC was consistently small for HR_peak (1-min: 4%, 3-min: 4%, incremental: 3%), moderate for absolute and body-mass normalized VO_2peak (1-min: 5%, 3-min: 6%, incremental: 7%) as well as PO_peak (1-min: 10%, 3-min: 9%, incremental: 6%) and large for BLa_peak (1-min: 20%, 3-min: 12%, incremental: 22%).

Fourteen and 9 participants changed their pacing strategy from day 1 to day 2 for the 1-min and the 3-min test, respectively. However, there were no differences in VO_2peak between day 1 and day 2 for neither the 1-min (1.1 ± 1.7 vs. 0.8 ± 1.7 mL·kg⁻¹·min⁻¹, p = 0.67) nor the 3-min test (0.7 ± 2.0 vs. 0.2 ± 2.5 mL·kg⁻¹·min⁻¹, p = 0.53) when comparing those who changed and those who maintained a stable pacing strategy across test days (Supplementary Figures 2, 3).

There was no significant interaction between type of test and test order on VO_2peak (p = 0.779). Furthermore, VO_2peak did not differ between test day 1 and 2 for the 3-min test (below 1% change, p > 0.05) or the incremental test (~2%, p = 0.068 and 0.085), but increased significantly for the 1-min test (~2%, p = 0.014 and 0.007). PO_peak was significantly higher on test day 1 as compared to test day 2 for the 1-, the 3-min and the incremental test (2, 1, and 4%, all p < 0.015). In line with the increased PO_peak, time to exhaustion significantly increased in the incremental test on day 2 (from 326 ± 63 to 346 ± 70 s, p = 0.003).

Our participants were highly trained for the poling movement and the exercise intensities used in this study, and they familiarized themselves with four times 5-min submaximal warm-up stages. We, therefore, chose to not perform a separate familiarization session for the peak tests in advance. However, in hindsight and as a recommendation for further studies, a separate familiarization session should be performed for all tests if the main outcome measure is PO_peak and for the 1-min and the incremental test if the main outcome measure is VO_2peak.

Furthermore, it remains to be investigated if other durations of the incremental test would result in different VO_2peak responses. As the participants in our study performed a thorough warm-up before starting the incremental test, we do not expect higher VO_2peak values with increases in duration of the test, but a follow-up study is needed to confirm this.

To be able to identify meaningful differences in body-mass normalized VO_2peak with paired comparison tests and similar participants in future studies, we estimated a sample size of 26 participants by n = 8·SDC²·(SEM²)⁻¹ as proposed by Hopkins (2000). Relatively similar numbers apply for most of the variables used in our approach. It is often challenging to recruit so many similarly upper-body trained participants, and in particular when aiming to test Paralympic athletes which are homogeneous with respect to their disability. Therefore, as large sample sizes as possible should be aimed at, if necessary through international collaborations. In addition, detailed description of the testing procedure and individual data should be made available so high-quality meta-analyses can be performed in the future.

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.