MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 1

1 A Maximal Rowing Ergometer Protocol to Predict Maximal Oxygen Uptake in

2 Female Rowers

4 Oscar B. Mazzaa, Søren Gama,b, Mikkel E. I. Kolinda,b, Christian Kiæra, Christina

5 Donstrupa, and Kurt Jensena*

a
6 Department of Sports Science and Clinical Biomechanics, University of Southern Denmark,
7 Campusvej 55, 5230, Odense M, Denmark
b
8 Department of Medicine, University hospital of Southern Denmark, Finsensgade 35, 6700,
9 Esbjerg, Denmark

10

11 Please cite as: Mazza, O. B., Gam, S., Kolind, M. E. I., Kiær, C., Donstrup, C. & Jensen, K. (2023).
12 A Maximal Rowing Ergometer Protocol to Predict Maximal Oxygen Uptake in Female Rowers.
13 SportRxiv.

14 Corresponding author:

15 *Kurt Jensen, Department of Sports Science and Clinical Biomechanics, University of Southern
16 Denmark, Campusvej 55, 5220, Odense M, Denmark

17 Tel. +45 65506089

18 Email: kjensen@health.sdu.dk
19

1
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 2

20 Abstract

21 Background: Laboratory assessment of maximal oxygen uptake (V̇O2max) is physically and

22 mentally draining for the athlete and requires expensive laboratory equipment. Indirect
23 measurement of V̇O2max could provide a practical alternative to laboratory testing. Purpose:
24 To examine the relationship between the maximal power output (MPO) in an individualized
25 7x2 min incremental test (INCR-test) and V̇O2max and to develop a regression equation to
26 predict V̇O2max from MPO in female rowers. Methods: 20 female club and Olympic rowers
27 (development group) performed the INCR-test on a Concept2 rowing ergometer to determine
28 V̇O2max and MPO. A linear regression analysis was used to develop a prediction of V̇O2max
29 from MPO. Cross validation analysis of the prediction equation was performed using an
30 independent sample of 10 female rowers (validation group). Results: A high correlation
31 coefficient (r=0.94) was found between MPO and V̇O2max. The following prediction
32 equation was developed: V̇O2max (mL·min-1) = 9.58*MPO (W) + 958. No difference was
33 found between the mean predicted V̇O2max in the INCR-test (3480 mL·min-1) and the
34 measured V̇O2max (3530 mL·min-1). Standard error of estimate was 162 mL·min-1 and %SEE
35 was 4.6%. The prediction model only including MPO, determined during the INCR-test,
36 explained 89% of the variability in V̇O2max. Conclusion: The INCR-test is a practical and
37 accessible alternative to laboratory testing of V̇O2max.

38 Keywords: indirect test; incremental test; maximal rowing performance test; V̇O2max;
39 prediction equation

40
41
42

2
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 3

43 Introduction
44
45 Although many physiological and anthropometric parameters influence performance in Olympic
46 rowing (2000 m), several studies have pointed towards maximal oxygen uptake (V̇O2max) as being
47 the most important physiological predictor of 2000 m (2k) rowing performance.1-4 Pripstein and
48 colleagues suggested that aerobic metabolic processes were responsible for 88% of total energy
49 production during a 2k ergometer race stimulation (2k-test) in female rowers.5 Despite rowing
50 competitions being faster nowadays, partly due to better equipment, the balance between anaerobic
51 and aerobic work would still assumed be approximately the same. As men and females compete at
52 the same distance (2k), females may, due to slower competition times, be even more dependent on
53 aerobic capacity compared to anaerobic capacity.
54 Therefore, regularly monitoring of rowers V̇O2max is important both in the selection
55 process for rowing teams and in the evaluation of training plans to ensure optimal adaption.6
56 Traditionally laboratory assessments of V̇O2max have been measured during a maximal 6 min or 2k
57 test performed on an air braked rowing ergometer.3,7,8 The 2k-test on ergometer mimics
58 performance during Olympic competitive rowing, where athletes row for 2000 m. Exercise time and
59 physiological demands are similar and ergometer rowing requires approximately the same
60 movement pattern as water rowing. The 2k-test has a high degree of reliability with a coefficient of
61 variance (CV) for mean power of less than 2%.7,8 This high degree of reliability makes it suitable
62 for measuring specific rowing performance in national and international indoor championships and
63 for team selection. However, the significant physiological and psychological stress from the 2k-test,
64 makes it unsuitable for regular monitoring aerobic performance changes. Anecdotal evidence
65 suggests that the test demands such a high level of exertion that it compromises motivation for
66 completing the tes,t and negatively impacts subsequent training sessions. Therefore, an indirect and
67 less demanding test that can be used frequently to estimate V̇O2max within relative narrow limits
68 would be desirable.
69 Several studies have suggested ways to indirectly assess V̇O2max in rowers with
70 varying degrees of success. Using a test consisting of submaximal rowing steps for 6 min at five
71 different incremental speeds and their corresponding heart rate (HR), Lakomy and Lakomy were
72 able to predict V̇O2max with a mean error of estimate <5%.9 Klusiewicz & Faff developed
73 regression formulas based on HR data from a submaximal test and the V̇O2max and average power
74 output measured during a 2k-test.10 These equations were able to estimate V̇O2max with between
75 5.1 and 7.5% total error in female rowers.10 However, sensitivity of the estimation was shown to be
76 too low to detect changes in V̇O2max during a training season.10 Kendall et al. proposed the use of
77 critical velocity and anaerobic rowing capacity to predict V̇O2max in female collegiate rowers.11
78 This method allowed for prediction of V̇O2max with a standard error of estimate (SEE) of 4.6%.11
79 This required the rowers to conduct four maximal efforts over a two-day period, which may be
80 difficult to implement in regular training schedules. Huntsman et al. developed a maximal
81 incremental test consisting of 7x2min steps with a 30 sec break.12 V̇O2 and Peak HR achieved at the
82 end of each step, were plotted in a linear regression model to predict V̇O2max. A moderate
83 correlation (r=0.55) was found in the men, while no correlation was found for the women,
84 suggesting the protocol was insufficient to reliably estimate V̇O2max in female and male rowers.12
85 In a recent study, Cherouveim et al 2022 found that maximal distance in the last step in a fixed 7
86 step incremental test and lean body mass, allowed for the estimation of V̇O2max with a CV of 3.3
87 and 2.1% in male and female adolescents, respectively.13 Recently, Jensen et al. proposed the use of
88 a continuous incremental test (INCR-test) with self-selected drag factor and stroke rate to predict
89 the V̇O2max of male rowers.14 The test used 7x2min steps with increasing intensity to exhaustion,
90 where starting and subsequent workloads were individually adjusted based on each rowers

3
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 4

91 estimated 2k performance.14 In the study, 20 male rowers were used to develop the prediction
92 equation and an independent sample of 14 male rowers were used for cross validation. The
93 developed prediction equation was able to reliably estimate V̇O2max with a 3.1% error (136
94 mL·min-1) using only maximal power output (MPO) as predictor variable.14
95 We hypothesized that the same test protocol conducted by Jensen and colleagues
96 would be able to predict V̇O2max in female rowers. It is therefore the aim of this study to develop
97 an equation that can be used to estimate V̇O2max in female rowers using the same INCR-test
98 method.
99

100 Methods

101
102 Participants

103 Thirty Danish female rowers (age 23.3 [2.84] years, height 1.74 [0.06] m, mass 70.0
104 [7.6] kg, fat free mass 49.8 [4.4] kg, rowing experience 5.2 [4.5] years), comprising 10 Olympic
105 rowers and 20 club and university rowers, volunteered to participate in the study after providing
106 written informed consent. Inclusion criterions were female rowers between 18 and 35 years with >6
107 months rowing experience and currently rowing ≥3 times per week (in boat or on ergometer).
108 Participants were excluded if they had performed strenuous exercise <24 hours before testing, or if
109 they were unwilling to pause administration of any ergogenic supplements (e.g., caffeine, creatine
110 etc.) prior to testing. Participants consented to participate in the test protocols and were informed of
111 all potential risks. The present study utilizes tests used during the ordinary training of the rowers.
112 As such, the present protocol did not require ethical approval according to the local ethical
113 committee (20212000-130).

114 Design

115 The participants performed the INCR-test on a Concept2 rowing ergometer

116 (Concept2, Model D, Morrisville, VT, USA) until exhaustion. To develop and to test the validity of
117 a prediction equation between MPO and V̇O2max, the participants were first balanced between
118 Olympic and club rowers in two groups (with equal proportions of Olympic and club rowers) and
119 then randomly assigned to a developmental group (n=20) and a validation group (n=10) (Table 1).
120 MPO data from the developmental group were used to develop a prediction equation for estimating
121 VO2 max.

122
123 Rowing ergometer tests

124 Before the INCR-test was performed, height was measured using a wall mounted
125 height scale. Weight and fat free mass were measured using a scale with bioelectrical impedance
126 analyzer (Tanita MC 780, Tanita Corp, Tokyo, Japan). The INCR-test was identical to the test used
127 by Jensen and colleagues.14 The rowers used self-selected stroke rate and drag factor during the test
128 and warm-up protocol. The initial step and the load increase per step in the INCR-test were
129 individually tailored to each rower’s performance level as suggested by Jensen.6 Following a brief
130 (5 min) warmup, participants rowed continuously, without pause, with intensity gradually
131 increasing every 2 minutes for 7± 1 steps or ~14 minutes. The starting step (step 1) equaled 40% of
132 the participants average power output during a 2k-test (W2k). For each subsequent step the

4
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 5

133 participants had to increase their power output by 10% of their personal W2k. Participants rowed
134 continuously for as many steps as possible until exhaustion. The power output in final steps
135 corresponding to 100 ± 10% of W2k. The test was stopped if the workload dropped by ≥10 Watts
136 from the prescribed workload for > 4 consecutive strokes. The MPO during the INCR test was
137 calculated as the average power at the last completed step plus 10% of their W2k multiplied by the
138 completed percentage of last initiated step. For example, if the participant had a W2k of 300 and
139 rowed for 13min (7 and a half step), their MPO would be 270 W + (50% of 30W) = 285 W. Test-
140 retest reliability data were obtained for MPO during two INCR tests performed at the same
141 weekday during two weeks of training in a group of 12 comparable female rowers. This resulted in
142 an interclass correlation of 0.99, a technical error of measurement of 3.3W (or 1.5%) and a CV of
143 1.4%.

144 Determination of V̇O2max

145 Oxygen uptake was measured based on a dynamic mixing chamber system (AMIS
146 Sport system; Innovision, Glamsbjerg, Denmark) for details see Jensen et al. 2021.14 HR were
147 recorded throughout the INCR-test using a HR monitor (Polar Sport Tester; Kempele, Finland).
148 Rate of perceived exhaustion (RPE) was recorded immediately after each test using a Borg Scale
149 (RPE 6-20). The highest mean 30-sec value for V̇O2 and Respiratory Exchange Rate (RER) during
150 the INCR test was recorded as V̇O2max and RERmax respectively. To further ensure that the
151 recorded value for V̇O2max represented a true maximum, 2 of the 3 following criteria had to be met
152 before the value for V̇O2max was accepted for the INCR-test: (1) RERmax > 1.10; (2) RPE rating ≥
153 17; (3)HR > 90% of the age predicted HRmax (Age predicted max= 220-age);
154
155 Statistical analyses
156
157 Data was analysed using Graph Pad Prism 7 software (Dotmatics, San Diego, USA).
158 Data from the developmental group was fitted in a linear regression model, using V̇O2max as the
159 dependent and MPO as the independent variable, to develop the V̇O2max prediction equation. A
160 stepwise regression model that included both MPO and fat free mass was also developed. Internal
161 cross validation analysis of the equation was conducted using the validation group. A parried T-test
162 was conducted to see if there was a statistically significant difference between measured and
163 predicted V̇O2max. SEE was calculated as the square root of 1/(n-2) multiplied by the sum of
164 residuals squared. Relative SEE (%SEE) was calculated as SEE/measured V̇O2max multiplied by
165 100. A Bland-Altman plot was created to observe if systematic bias was present. Results with p <
166 0.05 were considered significant. All data are presented as mean (SD) unless otherwise stated.
167
168 Results
169
170 The developmental group had an exercise time to exhaustion during the INCR-test of
171 843 (57) s or 7.0 ± 0.4 steps.Fitting data from the developmental group in a linear regression model
172 resulted in the following regression equation:
173
174 V̇O2max (mL·min-1) = 9.58*MPO + 958
175
176 A strong relationship between predicted and measured V̇O2max was observed using
177 data from the validation group (r=.97, P<.0001) (Figure 1). Accordingly, the prediction model
178 explained 89% of the variability in V̇O2max. Predicted V̇O2max was 3480 mL·min-1 whilst

5
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 6

179 measured V̇O2max was 3530 mL·min-1. No significant differences were observed between
180 predicted V̇O2max and measured V̇O2max (p=.2738). SEE was 162 mL·min-1 whilst %SEE was
181 4.6%. No drift in gas sensors measured before and after tests was observed in O2, while a minor
182 increase of 0.01% was seen for CO2 (p<.0001, Table 2). Using a stepwise regression model that
183 included both MPO and fat free mass was also developed. Adding fat free mass only improved the
184 prediction equation minimally (r=0,98 P=0.33).
185
186 ***Table 1 and 2 about here
187 ***Figure 1 about here
188
189 Discussion
190
191 In this study, we used the previously developed INCR-test, which was individualized
192 based on each participant W2k. Using this protocol, participants could work at approximately the
193 same gradually increasing relative intensity and reached exhaustion at approximately the same time.
194 No significant difference was found between measured V̇O2max (3530 mL·min-1) and
195 estimated V̇O2max (3480 mL·min-1) using the developed prediction equation (P=.2738). The
196 V̇O2max was predicted with a SEE of 162 mL·min-1 and %SEE of 4.6%, indicating that VO2max
197 obtained from the INCR test can accurately be predicted from MPO obtained from the same test.
198 In a recent study, Cherouveim et al 2022 measured performance during an incremental
199 7-step row test to predict the V̇O2max, of adolescent boys and girls between 13 and 17 years old
200 from a national development team .15 The authors were able to estimate V̇O2max with a CV of 3.3
201 and 2.1% for males and females, respectively.15 Unlike the present study, Cherouveim et al used a
202 non-individualized test procedure with a fixed increase of work output each step, intermittent rest
203 periods and gender-specific stroke rates. Cherouveim used body composition and test performance
204 to predict V̇O2max.15 Unlike Cherouveim, we did not see any benefit of adding any measure of
205 body composition to the regression model, as adding this parameter only improved the prediction
206 equation minimally (r=0,98 P=0.33 for lean mass). This may indicate that body composition is an
207 important factor to consider when dealing with adolescents . Club level rowers rarely have access to
208 valid measurements of body compositions and as such the inclusion of this would limit the practical
209 application of the prediction equation.
210 Kendall et. al. showed that V̇O2max could be predicted with a SEE of 144 mL·min-1 or 4.6% based
211 on critical velocity and anaerobic rowing capacity in female college rowers.11 This approach
212 required four exhaustive tests over 2 separate days on a rowing ergometer (400, 600, 800 and 1000
213 m).11 In comparison, the INCR-test is much less time consuming (~14 minutes) and induces less
214 physical stress on the rowers. Both Kendals et al.’s prediction equation and the one developed in the
215 present study have the same %SEE of 4.6%.
216 In another study, using the INCR-test, Jensen showed that the V̇O2max of male rowers could be
217 estimated using a 2k-test with almost the same level of accuracy as the INCR-test.14 However,
218 compared to the 2k-test, the INCR-test induced lower post-test blood lactate values and less fatigue,
219 indicating that the INCR-test would be more convenient for regular application than the 2k-test.14
220 Importantly, no statistically significant difference was observed between V̇O2max measured in the
221 INCR-test vs the 2k-test.14 Applying the prediction equation for the male rowers to the female
222 validation group from the present study, resulted in a significant difference between predicted and
223 measured V̇O2max values (p=<0.0001). This highlights the need for gender specific prediction
224 equations.
225 Changes in ambient conditions in the laboratory room during the exercise could
226 potentially influence results.16-18 In this study all tests were performed in well ventilated rooms to

6
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 7

227 secure stable ambient conditions. Electronic drift in the equipment could also lead to measurement
228 error.16 In this study, a small but statistically significant increase was observed in CO2 but not in O2
229 from pre- to post-test gas check (Table 2). In a few individual cases, where a drift in O2 larger than
230 0.05% was observed, the results were recalculated using the observed drift as reference.
231 Accordingly, in this study, the highest potential individual V̇O2max measurement error, due to gas
232 drift in O2 concentration would be less than 0.8%.
233 Several of the participants had no prior experience with the INCR-test. However, all were
234 accustomed to the Concept II row ergometer, and all received thorough verbal instructions and got
235 experience with the initial two levels of the INCR-test during warm-up. Test-retest variation in had
236 a CV of 1.4% in a comparable group of 12 female rowers, which was slightly lower that the
237 repeatability for male rowers.14. Future research is needed to determine the responsiveness of the
238 INCR-test to changes in V̇O2max during different periods over a rowing season. Our hypothesis
239 would be that the sensibility of the equation will decrease when changing training towards larger
240 portions of anaerobic training. We suggest testing MPO (by INCR-test) and measuring V̇O2max
241 before and after a training phase. Changes of each parameter could then be compared to determine
242 responsiveness.
243
244 Practical Applications
245
246 Compared to other available row performance test to estimate V̇O2max the INCR-test
247 may be more suited for testing the aerobic capacity of inexperienced or frail athletes since the tests
248 is based on individual performance. While the test requires an initial estimate of 2k-test
249 performance this is easily sidestepped in practice by giving a best estimate and having the athlete
250 row until exhaustion (as defined in present protocol). After which, the last achieved step is used for
251 future reference. Using this method, a single trial is, in our experience, sufficient to design a
252 progression schedule with participants reaching exhaustion at approximately step 7.
253 For regular monitoring of training status in female rowers, we suggest using the INCR-test, as this
254 test allows for athletes to be tested with approximately the same exercise time to exhaustion, the
255 same relative starting level and same relative intensity increment increase per step. From the
256 obtained MPO in the test, the V̇O2max can be estimated with an accuracy of ± 4.6%.
257
258 Conclusions
259
260 We have shown that MPO in the INCR-test can be used to accurately predict V̇O2max
261 in female club and elite rowers. Additional studies over longer periods of training are required to
262 test the responsiveness and accuracy of the V̇O2max predicted from MPO
263
264 Acknowledgements
265
266 The author thanks the rowers and coaches for their participation in this study. There
267 are no conflicts of interest. There are no companies or manufacturers who will benefit from the
268 results of this study.
269
270 References
271
272 1. Bourdin M, Lacour JR, Imbert C, Messonnier LA. Factors of Rowing Ergometer
273 Performance in High-Level Female Rowers. Int J Sports Med. 2017;38(13):1023-1028.

7
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 8

274 2. Cosgrove MJ, Wilson J, Watt D, Grant SF. The relationship between selected physiological
275 variables of rowers and rowing performance as determined by a 2000 m ergometer test. J
276 Sports Sci. 1999;17(11):845-852.
277 3. Ingham SA, Whyte GP, Jones K, Nevill AM. Determinants of 2,000 m rowing ergometer
278 performance in elite rowers. Eur J Appl Physiol. 2002;88(3):243-246.
279 4. Nevill AM, Allen SV, Ingham SA. Modelling the determinants of 2000 m rowing ergometer
280 performance: a proportional, curvilinear allometric approach. Scand J Med Sci Sports.
281 2011;21(1):73-78.
282 5. Pripstein LP, Rhodes EC, McKenzie DC, Coutts KD. Aerobic and anaerobic energy during
283 a 2-km race simulation in female rowers. Eur J Appl Physiol Occup Physiol.
284 1999;79(6):491-494.
285 6. Jensen K. Performance Assessment In: Secher NH, Volianitis S, eds. Rowing Blackwell
286 Publishing 2007:96-102.
287 7. Schabort EJ, Hawley JA, Hopkins WG, Blum H. High reliability of performance of well-
288 trained rowers on a rowing ergometer. J Sports Sci. 1999;17(8):627-632.
289 8. Smith TB, Hopkins WG. Measures of rowing performance. Sports Med. 2012;42(4):343-
290 358.
291 9. Lakomy HK, Lakomy J. Estimation of maximum oxygen uptake from submaximal exercise
292 on a Concept II rowing ergometer. J Sports Sci. 1993;11(3):227-232.
293 10. Klusiewicz A, Borkowski L, Sitkowski D, Burkhard-Jagodzińska K, Szczepańska B,
294 Ładyga M. Indirect Methods of Assessing Maximal Oxygen Uptake in Rowers: Practical
295 Implications for Evaluating Physical Fitness in a Training Cycle. J Hum Kinet.
296 2016;50:187-194.
297 11. Kendall KL, Fukuda DH, Smith AE, Cramer JT, Stout JR. Predicting maximal aerobic
298 capacity (VO2max) from the critical velocity test in female collegiate rowers. J Strength
299 Cond Res. 2012;26(3):733-738.
300 12. Huntsman HD, DiPietro L, Drury DG, Miller TA. Development of a rowing-specific
301 VO2max field test. J Strength Cond Res. 2011;25(6):1774-1779.
302 13. Almquist NW, Nygaard H, Vegge G, Hammarström D, Ellefsen S, Rønnestad BR. Systemic
303 and muscular responses to effort-matched short intervals and long intervals in elite cyclists.
304 Scand J Med Sci Sports. 2020;30(7):1140-1150.
305 14. Jensen K, Frydkjær M, Jensen NMB, Bannerholt LM, Gam S. A Maximal Rowing
306 Ergometer Protocol to Predict Maximal Oxygen Uptake. Int J Sports Physiol Perform.
307 2021;16(3):382-386.
308 15. Cherouveim E, Methenitis S, Simeonidis T, et al. Validity and Reliability of New Equations
309 for the Prediction of Maximal Oxygen Uptake in Male and Female Elite Adolescent Rowers.
310 Journal of Human Kinetics. 2022;83:77-86.
311 16. Gore CJ, Tanner RK, Fuller KL, Stanef T. Determination of Maximal Oxygen Consumption
312 (VO2max) In: Gore CJ, Tanner RK, eds. Physiological Tests for Elite Athletes Second
313 Edition Second Edition ed.: Australian Institute of Sport 2013:103-122.
314 17. Macfarlane DJ. Automated metabolic gas analysis systems: a review. Sports Med.
315 2001;31(12):841-861.
316 18. Shephard R. Open-circuit respirometry: a brief historical review of the use of Douglas bags
317 and chemical analyzers. Eur J Appl Physiol. 2017;117:1-7.
318
319

8
 MANUSCRIPT UNDER REVIEW PREPRINT Last updated 4/18/2023 9

320
321 Figure 1 - Linear relationship (A) and limits of agreement (B) between observed V̇O2max and the predicted V̇O2max using INCR test. Bias-lines
322 represent the mean difference between observed and predicted V̇O2max values. Dashed lines represent the 95% limits of agreement.
323
324
325
326
327
328
329
330 Table 1: Physiological Characteristics for Development and Validation group (Mean (SD))
Development Group n=20 Validation Group n=10
Age (y) 23.3 (3.35) 23.3 (1.49)
Height (m) 1.74 (0.06) 1.74 (0.05)
Weight (kg) 69.9 (7.65) 70.2 (7.99)
Experience (y) 5.2 (4.44) 5.2 (4.72)
V̇O2max (L/min) 3.4 (0.56) 3.5 (0.60)

331 Characteristics of both the development and validation group

332

333
334
Table 2: Drift in O2 and CO2 sensors before and after tests (Mean (SD))
Pretest, Posttest, Difference (P)
Mean (SD) Mean (SD)

Sensor O2, % 15.36 (0.52) 15.37 (0.51) 0.01 (0.587)

Sensor CO2, % 4.78 (0.55) 4.79 (0.55) 0.01 (0.0001)

335 Pre- and post-test values for sensor O2 – and CO2. No significant difference was found in O2, whilst a small statistically significant difference was
336 found in CO2
337