General characteristics of the subjects in all three groups are summarized in Table 1. Subjects in the three groups were similarly experienced cyclists, were of approximately the same age, had spirometry values that were within the normal range, and were predominately male. There were no significant differences in VO₂ max between the groups, either before or after training (Table 2).

The coefficient of variation for the time trial test was 3.94 % (Fig. 1, top panel), even though there were two subjects (one from the true control and one from the placebo training group) with vastly different values before and after the training period. If these two data points are not used, the coefficient of variation was only 0.78 %. In contrast, the coefficient of variation on the constant work-rate exercise test averaged 15.6 % (Fig. 1, bottom panel).

As expected, respiratory muscle strength, as assessed by measuring maximal inspiratory and expiratory mouth pressures, was unchanged by RMET. For example, peak inspiratory mouth pressure averaged 83.5 ± 8.3 cmH₂O before training, and 93.9 ± 7.2 cmH₂O after training, and peak expiratory pressure was 91 ± 10 cmH₂O before and 88.5 ± 11 cmH₂O after training. For the Control/Placebo group inspiratory pressure averaged 86.5 ± 7.4 cmH₂O before training, and 88.8 ± 7 cmH₂O after training; corresponding values for expiratory pressure were 104 ± 8.4 cmH₂O before training, and 97 ± 9.3 cmH₂O at the end of the training period. The maximal mouth pressure values that we recorded during both inspiration and expiration are similar to those observed in other young, fit subjects 131523, and none of the values changed significantly with training of the respiratory muscles, nor were there any differences between the subject groups. In contrast to strength training, the endurance capacity of the respiratory muscles (estimated as the sustainable ventilatory capacity) increased significantly after training in the RMET group, but not in any of the control or placebo group subjects (Fig. 2).

We assessed cycling performance by using the time trial test, and the performance of all subjects before and after the intervention period is shown in Fig. 3. With one exception, all subjects that underwent RMET improved their performance time (top panel). The average performance time fell from 47.1 ± 5.5 min before RMET, to 44.9 ± 5.5 min after RMET (Table 2, P < 0.05), a reduction of 4.75 %. Performance time did not change significantly in the control/placebo group (Table 2 and Figure 3, bottom panel). The percent change in performance time in the RMET group (-4.75 ± 1.6 %) and in the control/placebo group (0.46 ± 1.8 %) revealed a statistically significant difference (P = 0.015, Kruskal-Wallis ANOVA on ranks). As shown in Table 2, there were no significant changes in constant work-rate endurance time before and after the training period in either group.

We measured cardiorespiratory responses to constant work-rate exercise over the last 30 sec of each 10 % epoch of endurance time, before and after the intervention period. Pulmonary ventilation was higher after training in the RMET group and unchanged in the control/placebo group (Fig. 4). The increased V_E in the RMET group was explained by a significant rise in f (P = 0.013) with no change in V_T. The partial pressure of end-tidal CO₂ (P_ETCO₂) was not significantly different after the training period in either group (Fig 5). VO₂ increased significantly in the RMET group, and did not change in the control/placebo group (Fig. 6). Heart rate was the same before and after the training period throughout the exercise bout in both groups. For example, at the point of exhaustion in the RMET group, heart rate was 173 ± 2 beats/min before training and 172 ± 5.2 beats/min after training. For the control/placebo group, the corresponding values were 178 ± 3.6 beats/min before the training period, and 170 ± 8 beats/min after (P > 0.05 in both groups). Perceived exertion for both breathing and leg muscle effort rose monotonically and significantly (P < 0.001) over the course of the time trial test, but there were no significant differences after the intervention period in either the RMET or the control/placebo groups (data not shown).

An example of the changes in V_E and VO₂ during the time trial tests, before and after the intervention period, are given in figure 7 for a representative control and RMET subject. In all RMET subjects, V_E and VO₂ were higher in the post-RMET time trial test compared to the pre-RMET time trial test (Table 3). There were no consistent changes observed in the control/placebo group (Table 3). Note that the pre-training V_E was lower in the RMET group because the subjects in this group were significantly smaller, and three of the four female subjects were in this group (Table 1).

The correlation between the change in V_E and the change in performance time on the time trial test for all subjects is shown in Fig. 8. The correlation coefficient was -0.522, and was statistically significant (P = 0.0183).

The major, new finding of this study is that twenty sessions of respiratory muscle endurance training significantly increased both respiratory muscle endurance capacity and bicycle time trial performance in nine of ten fit, experienced cyclists and/or triathletes. We also showed that the improved performance was accompanied by an enhanced ventilatory response during constant work-rate exercise. We did not observe significant changes in either respiratory muscle endurance or time trial performance in a control group consisting of six subjects that did not train their respiratory muscles and four "placebo" subjects that trained their muscles at a low and ineffective intensity and duration.

Subjects that underwent RMET showed a significant, 12 % increase in the sustainable ventilatory capacity, with no change in respiratory muscle strength, demonstrating that the training protocol specifically enhanced respiratory muscle endurance performance. In contrast, the control and placebo groups did not show significant changes in either sustainable ventilatory capacity or respiratory muscle strength. The sustainable ventilatory capacity test was initially used by Leith and Bradley 23, who showed a 15 % increase in sustainable ventilatory capacity in subjects that underwent a respiratory muscle training program similar to that used in the present study. Fairbarn et al 10 also used the sustainable ventilatory capacity as an index of respiratory muscle training, and showed a 12% increase after sixteen endurance training sessions. Thus, the present results are consistent with those obtained by other investigators that have used the same training protocol. In contrast, Sonetti et al. 13 failed to show a significant improvement in the sustainable ventilatory capacity following five weeks of respiratory muscle training in fit cyclists, although they used a slightly different method than ours.

Previous studies have examined the influence of respiratory muscle endurance or strength training on VO₂ max and found no significant improvement 91018. However, VO₂ max is unlikely to be limited by an inadequate ventilatory response 24, so these negative results are not surprising. More recent studies have examined the influence of RMET on exercise endurance performance, and have shown that specific training of the respiratory muscles is associated with enhanced endurance performance, at least when the exercise tests require the subjects to work at about 70–80 % of their maximal capacity or less 2345678. Others have examined the effects of respiratory muscle training on the ability to perform very intense endurance exercise (85–95 % of maximal capacity), and uniformly find no change in performance 9101112. However, in all of these studies the subjects completed an open-ended exercise test, where they were asked to perform constant work rate exercise until "exhaustion". The subjective end point of such tests results in high variability, with the coefficient of variation for repeat testing in the same subjects ranging from 15 % (this study, Fig. 1) to as high as 30 % 1314. In contrast, the fixed end-point performance test, or "time trial" that we used has a coefficient of variation of 3.89 %, which is similar to recent work reporting values ranging form 1 % 13 to 3.4 % 14. Moreover, the time trial test is very similar to an actual bicycle race, wherein the goal is to cover the race distance, from starting line to finish line, in as short a time as possible.

Of five recent studies that used time trial tests to examine the influence of respiratory muscle training on exercise performance, three showed enhanced performance (on the order of 3 %) 151617, and two failed to find changes in performance that were significantly different than the changes observed in sham-training groups 1318. However, the three studies that did report enhanced time trial performance after training of the respiratory muscles 151617 were carried out by the same group; and more importantly, the respiratory muscles of their subjects were trained for strength, not endurance. Thus, the present study is the first to show that specific endurance training of the respiratory muscles leads to enhanced time trial performance in fit, young cyclists. Nevertheless, our findings are consistent with the results of recent studies that used strength rather than endurance training of the respiratory muscles 151617, indicating that either method can lead to enhanced exercise performance.

The almost 5 % improvement in cycling performance in the trained subjects is remarkable considering that our subjects were fit and already close to the limit of their performance potential. Thus, a race that a cyclist could complete in 40 minutes before respiratory muscle training would be completed in just over 38 minutes with a 4.7 % improvement in performance. We are unaware of other training methods that result in similar performance increments in experienced bicycle racers. Whether or not more extended or more intense respiratory muscle training would lead to even larger improvements in exercise performance is uncertain, but the results of this study suggest that this should be addressed.

Comparison of cardiorespiratory responses to CWE before and after RMET was done to determine if any key cardiorespiratory variables changed in response to the training stimulus. We found that V_E and VO₂ were significantly higher in CWE after RMET. Our subjects kept their training regimen constant, with the exception of adding the RMET, as evidenced by regular evaluation of their daily exercise logs. Thus, changes in the ventilatory response to CWE after RMET may be attributed to alterations in the performance of the respiratory muscles.

Accordingly, we examined the results of fifteen studies that reported ventilatory responses to exercise before and after a period of respiratory muscle training. Of the ten studies that used a respiratory muscle endurance training protocol, two report that V_E during exercise was reduced after training 23, four show no change in V_E 471013 two showed a significant increase 68 and two showed large average increases that apparently were not significant. For example, in the study by Morgan et al 9 V_E averaged 160 L/min during maximal exercise before training and 175 L/min after training, while McMahon et al 5 reported that sub maximal exercise V_E rose from 147 L/min before training to 156 L/min after training. Of the five studies that used a protocol designed to enhance respiratory muscle strength and power, four showed no change in exercise V_E after training 11121518, and one showed an increase that just failed to reach significance (P= 0.051, 17. Our data demonstrate a consistently higher ventilatory response to both constant work rate exercise and a simulated time trial test after RMET. It is important to emphasize that the increase in V_E was observed in all ten subjects in the RMET group. Thus, our data are consistent with the results of previous studies, wherein the results of all but two of fifteen show either no change in V_E, a significant increase in V_E with RMET, or a clear trend towards an increase in V_E.

Although we cannot explain the rise in V_E during constant work-rate exercise after RMET, we found a significant, inverse correlation between the change in V_E and the change in performance time in all 20 subjects, with the subjects that ventilated more after the training period having better performance. This is at odds with the data of Boutellier et al 2, who showed the opposite effect (see their Fig. 4). Given this qualitative difference, we interpolated all of the numbers from their figure 4 and re-ran the data analysis. Several interesting observations result: first, they studied 8 subjects, but ran the correlation between V_E and cycling performance time on only 7 because "minute ventilation in subject AH showed such an unrealistically large scatter during the endurance cycling test after respiratory training that the values had to be disregarded"; second, when we analyzed the data we found a similar correlation coefficient (0.75), with a P value of 0.051, but noticed that the data were dominated by a clear outlier that had a 156 % increase in cycling performance and a 35 % decline in V_E after training. When we removed this outlier from the analysis, the correlation coefficient fell to 0.45, and the P value rose to 0.364. Thus, the relation that they reported was dominated by the data from this one subject. In a more recent report from the same laboratory 6, 6 of 8 subjects that had higher exercise V_E after RMET had reduced cycling endurance performance, although statistics were not reported in that analysis. We cannot explain the reason behind the discrepant findings, but wish to emphasize that our data are consistent and statistically significant, and that more detailed studies of this issue are needed.

The increase in V_E observed during exercise after RMET was associated with a higher whole body VO₂, suggesting that the respiratory muscles consumed at least some of the extra oxygen to cover the requirements of the increased ventilatory response. This would appear to "steal" more blood from the exercising locomotor muscles (see 19 for discussion) and therefore predispose the subject to an earlier onset of fatigue 20. However, our trained subjects performed better, with the performance enhancement associated with the increase in post-training V_E. This study was designed to examine the change in performance after RMET (see above), and because of this we avoided invasive measures that could have provided insight into mechanisms. For example, it is possible that the higher ventilatory response to exercise after RMET improved oxygen transport and/or acid base balance during exercise 25. However, Stuessi et al 4 measured arterial blood gases during constant work-rate exercise before and after respiratory muscle training and found no differences in acid base or oxygen transport variables. However, they did not observe changes in ventilation with training, so blood gases and pH would not be expected to change significantly.

A possible explanation for our findings is that RMET allowed the subjects to incur a higher ventilatory load without any increase in dyspneic sensations. Although a higher ventilatory response during exercise might be expected to lead to greater feelings of dyspnea, we found no significant changes in sensations of dyspnea or leg effort after RMET. Thus, in spite of the higher V_E after RMET, the subject's perception of dyspnea (and leg fatigue) was unchanged. In other words, the subjects could tolerate a greater ventilatory load without an increase in dyspneic sensations after RMET, which may lead to enhanced exercise performance.

In conclusion, twenty sessions of rigorous endurance training of the respiratory muscles lead to significantly improved exercise performance (on average, 4.7 %) in fit, experienced cyclists and/or triathletes. The improved cycling performance after the training period was significantly correlated with an increase in the ventilatory response to exercise, without an increase in sensations of dyspnea; thus, the subjects could breathe more without increased feelings of breathlessness. The mechanism behind either the enhanced performance or the enhanced ventilatory response after respiratory muscle endurance training remains to be established.

Twenty fit, experienced cyclists and/or triathletes agreed to participate in the study. Requirements for enrollment included that they: (a) be a cyclist or triathlete, (b) be involved in at least 3 hours of endurance training per week, (c) keep a constant level of training over an 8 week period and (d) be free of any past or present respiratory illnesses. Using a standard randomization procedure, we initially assigned 10 subjects (3 female, 7 male) to the RMET group and 6 to a control group (1 female, 5 male). We subsequently added 4 additional subjects (4 males) that did "placebo training" (see below). The Human Subjects Committee at the University of Arizona approved the study protocol, and all subjects provided written informed consent prior to participation.

The study protocol was divided into three phases:

Forced Vital Capacity (FVC) and Maximal Voluntary Ventilation in 12 seconds (MVV₁₂) were measured on a spirometer (model 827, Ohio Medical, Toledo, OH, USA). Forced Expiratory Volume in 1 second (FEV₁) was obtained from the FVC tracing to calculate the FEV₁/FVC ratio. For the MVV₁₂, subjects were instructed to "move as much air as possible" and were given feedback from the investigator until a maximal level of ventilation was obtained. For both tests, the best two out of three trials were averaged, and expressed under BTPS conditions as described previously 26.

To assess the endurance of the respiratory muscles, we used a SVC test similar to that used by Leith and Bradly 23. A breathing mask and head cap (Hans Rudolph, Kansas City, MO, USA) were secured to the subject, and to prevent air leakage, the mask was sealed to the bridge of the nose with dental impression material (Exaflex, GC America, Alsip, Ill, USA). A large 2-way valve (Hans Rudolph 3700) was attached to the front of the mask. The subjects inspired air from a Douglass bag that was partially filled with warm water to prevent throat dryness during the test. A mixture of 10% CO₂, 21% O₂ (nitrogen balance) from a separate gas tank was fed into the Douglass bag via flexible tubing for maintenance of end-tidal CO₂ (ETCO₂) throughout the sustained period of hyperpnea. Respiratory tubing connected the bag to the inspired side of the breathing valve and was interrupted by a pneumotachometer (model 3813, Hans Rudolph) connected to a differential pressure transducer (Validyne MP45-871, Irvine, CA, USA) to measure inspired flow. The flow signal was split and sent to a polygraph integrator (model 7P10F, Grass, Quincy, MA) and to an analog-to-digital converter (model 1401 plus, Cambridge Electronic Design, London, UK). The integrated flow signal was sent to a chart recorder (model TA11, Gould, Courtaboeuf, France), and the raw flow signal was recorded digitally (Spike II software, Cambridge Electronic Design) for later analysis. To ensure the linearity and accuracy of the system, a 3-liter calibration syringe (Hans Rudolph) was used to calibrate the integrated flow signal before each test, while the raw flow signal was calibrated over a range extending from 0–220 L/min with a Matheson Rotameter. The expired side of the breathing valve was open to the room. A small sample of expired air was collected from each expired breath and sent to a rapidly responding CO₂ analyzer (model CD-3A, Ametek, Pittsburgh, PA, USA). The CO₂ signal was sent to the chart recorder such that the investigator could continuously monitor end-tidal CO₂. A one-two minute period of room air breathing was necessary prior to each test to find the resting end-tidal CO₂ level. Once established, this level of end-tidal CO₂ was maintained during all breathing maneuvers by addition of variable amounts of the air/CO₂ mixture to the Douglass bag.

A three-minute warm up period (at approximately 80% MVV) preceded all tests. Before the test began, the subjects were reminded to breathe maximally ("like the MVV") for "at least 10 minutes". Although measurements of V_E were made each minute over the first six minutes of the test (see Fig. 3), the extra few minutes of hyperpnea allowed us to insure that a clear ventilatory asymptote was reached in all subjects. The first test was considered a practice test and the subject was given immediate feedback on the results. At least two more tests followed on separate days (see below). In order for a test to be included in the final analysis, V_E during the first 30 breaths of the test had to be greater than or equal to the subjects MVV to ensure that the effort was maximal. In addition, V_E had to decrease monotonically over the course of the test. Further analysis of each test consisted of finding the average ventilation over 30 breaths occurring at the beginning of each minute of the first six-minutes of the test (Fig. 3). The average V_E during the plateau phase (over the last 2–3 minutes of the test) was taken as the SVC. The SVC from the two tests had to be within 10% of each other before the training phase commenced, and this was considered a baseline measurement to which all subsequent SVC tests were compared.

Inspiratory and expiratory airflow rates were measured breath-by-breath with a turbine flowmeter attached to a facemask that allowed the subjects to breathe orally or nasally (Vacumed, Torrance, CA, USA). This configuration is designed for very low dynamic flow resistance, which we calculated as 0.25 cmH₂O/L/sec at a flow of 6 L/sec. A small bore tube connected to a vacuum pump continuously sampled gases in the mask at a flow rate of 200 ml/min, and the fractional concentrations of CO₂ (F_ETCO₂) and O₂ (F_ETO₂) were measured with Vacumed analyzers (Vista mini-CPX Silver Edition, VacuMed). End-tidal CO₂ was monitored with a rapidly responding CO₂ analyzer (model CD-3A, Ametek) that sampled gas via a second port in the facemask. Gas analyzers were calibrated with precision gas and the flow meter was calibrated with a precision 3.0 L syringe (W.E. Collins, Inc., Braintree, MA). Heart rate (HR) was monitored and recorded on-line with a HR interface unit (VacuMed).

Subjects performed a progressive intensity exercise test to exhaustion on an electronically braked cycle ergometer (Excalibur, Lode, Stockholm, Sweden). The first 15 minutes of the test consisted of sitting motionless for 5 minutes (to collect resting data), followed by a 10-minute warm-up period. Upon completion of the warm-up, the workload was set at 50 W and increased to 100 W after 1 minute, and then by 30 W (25 W for women) every minute until a pedal rate of at least 60 revolutions per minute could no longer be maintained. A test was considered maximal if two of the following three criteria were reached: the respiratory exchange ratio (RER) reached 1.10, HR reached the age-predicted maximum, or VO₂ plateaued or decreased with an increase in workload.

To compare ventilatory and metabolic parameters at the same work-rate before and after RMET, subjects performed a constant work-rate exercise test to the point of volitional exhaustion. The work-rate for this test was set at approximately 80 % VO₂ max. Subjects warmed up for 10 minutes at the wattage of their choice, and then the wattage was ramped up to the target work-rate within 2 minutes. Endurance time was the elapsed time between reaching the target work-rate and when the subject could no longer maintain a pedal rate of 60 revolutions per minute or higher, as described previously 26. Parameters measured and displayed every six breaths included V_E (BTPS), tidal volume (V_T) (BTPS), respiratory rate (f), HR, VO₂, carbon dioxide production (VCO₂), RER, and ETCO₂.

To test cycling performance while minimizing the variability observed in open-ended cycling endurance tests (1314, and see Fig. 1), subjects performed a cycling task meant to simulate a time trial. When set in linear mode, the Lode ergometer becomes pedal rate dependent, such that as pedal speed increases, work rate also increases according to the following formula:

W = L·(RPM)²

Where L is a constant linear factor and RPM is the pedaling rate in revolutions per minute. The linear factor L was chosen so that each subjects preferred pedaling rate (observed during constant workload exercise) would cause them to work at the same wattage that elicited approximately 80 % VO₂ max. Then, the target amount of work each subject had to complete was calculated based on the following formula:

Target amount of work (J) = 0.80·Wmax·2700 (seconds)

Where Wmax was the maximal workload achieved during the incremental cycling test and 2700 seconds (45 minutes) was chosen to simulate a 40-kilometer time trial. The subject was instructed to complete the required work as quickly as possible. The subject viewed the required work on a computer monitor, and as work was done the number decreased accordingly, with the test ending when the required work was completed (i.e., a reading of zero on the computer monitor). The subject was unaware of the elapsed time, the work-rate, and the HR. Subjects were instructed to warm up for 10 minutes at the wattage of their choice before the test began. Performance time was the time elapsed between the onset of the time trial and attainment of the target (i.e., work-rate = 0).

To determine the strength of the respiratory musculature, maximal inspiratory and expiratory mouth pressures were measured, as described in detail previously 26. Maximal mouth pressure was measured from a port in a mouthpiece that was connected via a small length of tubing to a pressure transducer and bridge amplifier (Coulbourn Instruments, Allentown, PA, USA). The mouthpiece contained a valve that could be manually opened and closed by the subject. There was a small leak in the valve to equalize pressure and to prevent buccal pouching. With nose clips attached and the valve open, the subject was instructed to either inhale (to total lung capacity) or exhale (to residual volume) maximally, and then to quickly close the valve and exhale or inhale in concert with a cadence produced by an investigator. The cadence was one that we have used previously when measuring upper airway or expiratory muscle force 2627, and is described as follows: "up, 1, 2, 3, hold, hold, relax". In this way the subjects had to reach their maximum pressure gradually, over 3–4 sec, and then hold it for another two seconds. Our previous work has demonstrated that this method is highly reproducible and well understood by the subjects. We then measured the average pressure over a window encompassing 1.5–3 sec into the maneuver as our peak mouth pressure. The pressure signal was sent to the chart recorder and displayed to the subject via the monitor of a digital oscilloscope. Three inspiratory and 3 expiratory maneuvers were done, and the best two out of three efforts were averaged.

Approximately every 2–3 min throughout the time trial and constant work-rate exercise tests, the subjects were asked to estimate their respiratory and leg effort using a visual analogue scale as described previously 26. The scale ranged from 0–20, and the subjects simply pointed to the appropriate number when asked.

Fourteen subjects completed twenty sessions of RMET over a 4-week period, a protocol that has been used successfully by others 2910. For the experimental subjects, each training session lasted 30 minutes. All subjects wore a mask (see above) with a two-way non-re-breathing valve from which the diaphragms were removed. A rubber cork was placed on the expired side of the valve such that the subject both inspired and expired from the inspiratory side of the valve. Respiratory tubing connected the inspired side of the valve to a pneumotachometer and pressure transducer (Validyne MP45) to measure airflow. The integrated flow signal was displayed on an oscilloscope monitor placed directly in front of the subject. As in the SVC test, a small sample of each expired breath was analyzed and sent to the chart recorder to monitor end-tidal CO₂. Tubing of various lengths could be attached to the distal side of the pneumotachometer to add sufficient dead space to maintain end-tidal CO₂ at the resting level.

Each day, a ventilatory target was set and the subject had to maintain the target for the duration of the training session. The target V_T was set on the oscilloscope screen. Breathing frequency was set with a metronome such that the subject inspired and expired to the sound of a tone. For the experimental group, we used an initial target that matched the subject's maximal V_T and f achieved during the incremental cycling test (94.8–134.5 L/min). Tidal volume or f was increased every one-two days for the first two weeks to continuously challenge the respiratory musculature. This was semi-quantified by keeping subjects at a level of ventilation that induced a value of 18–19 on the 20 point respiratory effort scale (see above).

For the second two weeks, V_T reached an upper limit while f continued to increase. Subjects were instructed to consider the target V_T and f as a goal rather than a limit. Many times, the subjects voluntarily increased V_T and f in order to increase the intensity of RMET. To further enhance the training, subjects were asked to perform a 30–60 second MVV three-four times during most training sessions. The placebo group followed the same protocol as the experimental group, except that each training session lasted only 5 minutes, the target ventilation rate was set at 65% of maximal V_T and breathing f, and was never increased. Subjects in the control group did no respiratory muscle training at any time. To avoid biasing the training effort, we did not verbally encourage either the training or placebo group during the RMET sessions.