The high correlation (r2 = .86-.96) and agreement between exercise and predicted VO2 support the validity of the model, which provides accurate VO2peak estimations after a single maximal swim while avoiding the error of backward extrapolation and allowing the subject to swim completely unimpeded.

To assess the validity of postexercise measurements to estimate oxygen uptake (V˙O) during swimming, we compared V˙O measured directly during an all-out 200-m swim with measurements estimated during 200-m and 400-m maximal tests using several methods, including a recent heart rate (HR)/V˙O modelling procedure. 25 elite swimmers performed a 200-m maximal swim where V˙O was measured using a swimming snorkel connected to a gas analyzer. The criterion variable was V˙O in the last 20 s of effort, which was compared with the following V˙O estimates: 1) first 20-s average; 2) linear backward extrapolation (BE) of the first 20 and 30 s, 3×20-s, 4×20-s, and 3×20-s or 4×20-s averages; 3) semilogarithmic BE at the same intervals; and 4) predicted V˙O using mathematical modelling of 0-20 s and 5-20 s during recovery. In 2 series of experiments, both of the HR/V˙O modelled values most accurately predicted the V˙O (mean ∆=0.1-1.6%). The BE methods overestimated the criterion values by 4-14%, and the single 20-s measurement technique yielded an underestimation of 3.4%. Our results confirm that the HR/V˙O modelling technique, used over a maximal 200-m or 400-m swim, is a valid and accurate procedure for assessing cardiorespiratory and metabolic fitness in competitive swimmers.

Prior reports have described the limitations of quantifying internal training loads using hear rate (HR)-based objective methods such as the training impulse (TRIMP) method, especially when high-intensity interval exercises are performed. A weakness of the TRIMP method is that it does not discriminate between exercise and rest periods, expressing both states into a single mean intensity value that could lead to an underestimate of training loads. This study was designed to compare Banister's original TRIMP method (1991) and a modified calculation procedure (TRIMPc) based on the cumulative sum of partial TRIMP, and to determine how each model relates to the session rating of perceived exertion (s-RPE), a HR-independent training load indicator. Over four weeks, 17 elite swimmers completed 328 pool training sessions. Mean HR for the full duration of a session and partial values for each 50 m of swimming distance and rest period were recorded to calculate the classic TRIMP and the proposed variant (TRIMPc). The s-RPE questionnaire was self-administered 30 minutes after each training session. Both TRIMPc and TRIMP measures strongly correlated with s-RPE scores (r = 0.724 and 0.702, respectively; P < 0.001). However, TRIMPc was ∼ 9% higher on average than TRIMP (117 ± 53 vs. 107 ± 47; P < 0.001), with proportionally greater inter-method difference with increasing workload intensity. Therefore, TRIMPc appears to be a more accurate and appropriate procedure for quantifying training load, particularly when monitoring interval training sessions, since it allows weighting both exercise and recovery intervals separately for the corresponding HR-derived intensity.

To assess the validity of postexercise measurements in estimating peak oxygen uptake (V̇O2peak) in swimming, we compared oxygen uptake (V̇O2) measurements during supramaximal exercise with various commonly adopted methods, including a recently developed heart rate - V̇O2 modelling procedure. Thirty-one elite swimmers performed a 200-m maximal swim where V̇O2 was measured breath-by-breath using a portable gas analyzer connected to a respiratory snorkel, 1 min before, during, and 3 min postexercise. V̇O2peak(-20-0) was the average of the last 20 s of effort. The following postexercise measures were compared: (i) first 20-s average (V̇O2peak(0-20)); (ii) linear backward extrapolation (BE) of the first 20 s (BE(20)), 30 s, and 3 × 20-, 4 × 20-, and 3 or 4 × 20-s averages; (iii) semilogarithmic BE at 20 s (LOG(20)) and at the other same time intervals as in linear BE; and (iv) predicted V̇O2peak using mathematical modelling (pV̇O2(0-20)]. Repeated-measures ANOVA and post-hoc Bonferroni tests compared V̇O2peak (criterion) and each estimated value. Pearson's coefficient of determination (r(2)) was used to assess correlation. Exercise V̇O2peak(-20-0) (mean ± SD 3531 ± 738 mL·min(-1)) was not different (p > 0.30) from pV̇O2(0-20) (3571 ± 735 mL·min(-1)), BE(20) (3617 ± 708 mL·min(-1)), or LOG(20) (3627 ± 746 mL·min(-1)). pV̇O2(0-20) was very strongly correlated with exercise V̇O2peak (r(2) = 0.962; p < 0.001), and showed a low standard error of the estimate (146 mL·min(-1), 4.1%) and the lowest mean difference (40 mL·min(-1); 1.1%). We confirm that the new modelling procedure based on postexercise V̇O2 and heart rate measurements is a valid and accurate procedure for estimating V̇O2peak in swimmers and avoids the estimation bias produced by other methods.