The concept and definitions of anaerobic and ventilatory thresholds (VT) have sparked considerable literature and debate over the years. One of the reasons for this may reside in the lack of a consensual definition of both concepts and the proliferation of terms used to describe them (34). The anaerobic threshold can be defined as the oxygen consumption above which aerobic metabolism is supplemented by anaerobic mechanisms, and after which, a progressive increase in blood lactate concentration and metabolic acidosis occur (39). The onset of this pivotal event can be estimated using noninvasive techniques based on the nonlinear evolution of carbon dioxide production (V˙CO2) and minute ventilation (V˙E) relative to V˙O2 during incremental exercise (the so-called VT). In particular, the breaking point in the V˙CO2–V˙O2 relation (V-slope method) (3) and the moment at which there is a rise in the ventilatory equivalent for O2 (V˙E/V˙O2) without a concomitant rise in ventilatory equivalent for V˙CO2 (V˙E/V˙CO2) (the ventilatory equivalent method (VEM)) (29) have been used to identify VT. Exercise above the VT is associated with reduced exercise tolerance, metabolic acidosis, a slowing of oxygen consumption (V˙O2) and CO2 production (V˙CO2) kinetics (38), and a sharp rise in dyspnea (43).
In the clinical setting, VT is used as a predictor of overall aerobic fitness and is responsive to training both in healthy subjects and in patients with chronic diseases (26,36). It is a prognostic marker in chronic cardiorespiratory diseases (16,22,27) and in the perioperative period (35,40). Because of its close relation with overall exercise tolerance, the VT is also useful as a tool for exercise prescription. Patients with chronic obstructive pulmonary disease (COPD) greatly benefit from exercise training, and pulmonary rehabilitation has become a standard of care in the management of these patients (33). Compared with healthy individuals, however, patients with COPD exhibit a marked reduction in exercise tolerance caused partly by expiratory flow limitation and dynamic hyperinflation, increased work of breathing, abnormal breathing pattern, high VD/VT ratio, and gas exchange anomalies (13,23) and as such may be unable to tolerate prolonged high-intensity training. Training programs using the VT as a tool to guide exercise intensity have been safely and successfully used in this population (8,15,31,36,37), and additional data suggest that such an “individualized” prescription may offer an advantage over an “interval-based” regimen (10,19,42).
Both the V-slope and the VEM rely on a manual manipulation by an observer or an automated computerized analysis and as such are prone to variation and error. The presence of ventilatory and gas exchange anomalies in patients with COPD may further impair the reliable identification of VT using these techniques. A large intra- or interobserver variation may have consequences when using the VT for exercise prescription, when monitoring response to training, or when performing prognostic evaluation in patients undergoing surgery or patients with heart failure that have concomitant COPD. In healthy subjects, the high intraobserver reliability of the measurement of the VT has already been demonstrated (17), but the interobserver reliability showed more heterogeneous results (12,14,17,41). In patients with COPD, one study (4) showed acceptable interobserver variability in the identification of the VT but was limited by its small sample size and the lack of details regarding the clinical characteristics of the patients included.
We hypothesized that COPD severity would negatively affect the interobserver reliability of the identification of the VT as determined from the V-slope method and VEM. In accordance, the aims of this study were 1) to quantify the reliability of human observers in determining VT in control subjects and patients with COPD, 2) to compare human versus computerized analyses of VT, and 3) to evaluate whether the interobserver difference in VT identification amounts to a clinically significant difference in the corresponding HR (HRVT).
This study was based on an analysis of incremental exercise test data from individuals who completed an exercise test in the respiratory physiology laboratory at l’Hôpital du Sacré-Coeur de Montréal. Data from all pulmonary function tests, exercise tests, and blood gas analyses performed since March 2010 were stored in a common database located on a standalone computer in the physiology laboratory. Data for both patients with COPD and controls were extracted from this database. The study was approved by the institutional ethics committee.
A convenience sample of individuals with COPD and controls was selected from the aforementioned database. For patients with COPD, inclusion criteria were as follows: age ≥ 40 yr old, history of smoking of at least 20 pack-years, an objective diagnosis of COPD (as assessed by clinical evaluation and a spirometry result showing a postbronchodilator forced expiratory volume in 1 s (FEV1)/forced vital capacity (FVC) ratio less than 0.70 and an exercise test duration time of at least 6 min. This last criterion was implemented to maximize the chance of observing a VT.
With the assumption that patients with Global initiative for chronic Obstructive Lung Disease (GOLD) stage 4 disease would be less represented in the database, they were selected first. The database was then screened to identify, for each very severe patient, a matching subject among all other severity groups and among controls. Matching was based on age (±4 yr), sex, and body mass index (BMI, ±4 kg·m−2). Control subjects were defined as individuals with normal resting pulmonary function tests, normal V˙O2peak (i.e., ≥85% V˙O2max predicted), and normal cardiorespiratory response to exercise and were matched to patients with COPD for age, sex, and BMI (see earlier data). Reasons for referral to a cardiopulmonary exercise test (CPET) in control patients were as follows: unexplained dyspnea on exertion (16 patients), preoperative evaluation (five patients), and lung cancer (two patients).
Subjects were excluded from the study if their medical file suggested clinical disease worsening or a respiratory exacerbation in the 4 wk preceding the exercise test, evidence of another condition that could limit exercise performance (asthma, unstable CHD, heart failure, cancer, symptomatic peripheral vascular disease, or significant osteoarthritis), long-term oxygen therapy, or incomplete baseline evaluations.
Twenty-three patients with GOLD 4 disease meeting inclusion criteria and having a suitable match in all other disease severity groups were identified and included in the study (total sample, 115 patients). On the basis of the results of the means and SD of the first 65 patients studied, a sample size of 22 patients per disease subgroup was necessary to identify a difference on 100 mL·min−1 in V˙O2VT between controls and each COPD group with a power on 80% and α-level of 0.05.
Demographic and clinical information were collected from medical files. These include age, sex, BMI, ethnicity, current medication, and self-reported smoking status. Lung function was assessed using spirometry (for expiratory flow rates), body plethysmography (for lung volumes), and single breath hold technique (for lung diffusion capacity for carbon monoxide). All tests were performed and interpreted according to American Thoracic Society guidelines in a laboratory at sea level.
Symptom-limited incremental exercise tests were performed according to published guidelines (1). More specifically, tests were realized on an electromagnetically braked cycle ergometer (Ergoline 200; Ergoline, Bitz, Germany) with a protocol including a 2-min rest and a 3-min initial unloaded cycling. Load was increased linearly until exhaustion (ramp was individually determined for each patient by the attending physician on the basis of either previous exercise testing result or expected maximal work rate as estimated by overall physical fitness and/or FEV1) with the goal of maintaining a cycling speed of 60 rpm. Breath-by-breath analysis of expired gases was performed using electronic analysis (Jaeger Oxycon Pro; CareFusion, Höchberg, Germany). V˙E, V˙O2, V˙CO2, V˙E/V˙O2, and V˙E/V˙CO2 were computed using 20-s averages of breath-by-breath values. Peak V˙O2 was the highest 20-s mean V˙O2 obtained. Patients using beta-blockers were not required to withhold them before performing CPET. Oxygen saturation was monitored using finger or ear pulse oximetry. Exercise capacity was defined as the highest work rate achieved for at least 20 s at a rate of at least 50 rpm. Arterial blood gases were assessed at baseline using a standard blood gas analyzer (ABL800 Flex; Radiometer, Copenhagen, Denmark). Dyspnea and leg fatigue were evaluated at rest and at maximal exercise intensity using the modified 10-point Borg scale (7). Additional information regarding internal quality control can be found in the “Methods” section of the supplemental digital content file (see Document, Supplemental Digital Content, Methods and Results, http://links.lww.com/MSS/A579).
For all patients, the following two manual methods were used to identify the VT: 1) the V-slope method and 2) the VEM. To optimize the validity of the V-slope method, care was taken to ensure that the ranges of V˙O2 and V˙CO2 in the plots were equal (3) and that the V˙O2 scale was adequate to allow a precise identification of V˙O2VT. In addition, a computer-generated analysis of VT (LABManager version 126.96.36.199; CardinalHealth, Höchberg, Germany) was used again using both the V-slope method and the VEM. Human observers took care to exclude the first minute of exercise from analysis to avoid confounding VT with a “pseudothreshold” (24) sometimes associated with hyperventilation at the onset of exercise. For all subjects, computerized determination of VT was allowed between the onset and the termination of incremental exercise; the warm-up and recovery periods were excluded from analysis by the software. In the event where a VT could not be identified, it was reported as “undetermined.”
The VT was reported both as the V˙O2 at which it occurred (in absolute value) and as the corresponding HR (HRVT). Graphs for VT analysis for both methods were extracted from the database by a research assistant unrelated to the study, coded, duplicated, and submitted to two observers (B. P. D. and M. M.), who blindly recorded the presence or absence of a VT, its value in milliliters of V˙O2, and the corresponding HR. The graphs used by the human observers were identical to the ones used for computerized analysis. Precise and identical instructions on how to identify VT using both the V-slope method and the VEM were given to both observers. For the V-slope method, VT was defined as “the breaking point in the line of the graphical representation of V˙CO2 against V˙O2” (3). For the VEM, VT was defined as “the point where V˙E/ V˙O2 begins to increase while V˙E/V˙CO2 remains stable, when both are plotted against V˙O2” (29).
To test for internal validity, a subsample of 50 graphs drawn randomly from COPD and controls was blindly resubmitted to the two observers for a second VT determination. Both observers were physicians with formal medical training in respiratory medicine, specific training in exercise physiology, but less than 5 yr of clinical experience. VT analyses were performed in an independent and blinded manner.
Agreement between the human observers in the determination of V˙O2VT was assessed using intraclass correlation coefficients (ICC) (2,1—two-way random single measure). Reliability using ICC was interpreted according to the following scale: virtually none for ICC ≤ 0.10, slight for ICC of 0.11–0.40, fair for ICC of 0.41–0.60, moderate for ICC of 0.61–0.80, and substantial for ICC ≥ 0.81 (32).
To test whether human and computer analyses of V˙O2VT are interchangeable, Bland–Altman graphical analysis and Passing–Bablok regression analysis were performed (25). This nonparametric statistical tool allows the estimation of the interchangeability of two analytical methods and of the possible bias between them. It provides a numerical quantification of agreement levels and does not make any assumption about the distributions of the samples of their measurement errors and is nonsensitive to outliers. It does however require that data be continuously distributed and linearly related.
The mean differences between the two human observers’ assessment of V˙O2VT were compared across disease category groups using one-way ANOVA with post hoc Bonferroni correction.
A stepwise multiple linear regression analysis that included baseline demographic data and pulmonary function and exercise test results was performed to identify independent predictors of a larger interobserver difference in V˙O2VT.
The mean interobserver difference in HRVT for each of the five subgroups was compared using one way ANOVA with post hoc Bonferroni correction. An empirical threshold of ±5 bpm (total range of 10 bpm) was chosen as the cutoff for clinical significance for this parameter because we believe that when exercise training is based on a target HR value, the training HR should stay within this limit of the objective.
Intraobserver reliability was assessed using ICC. All analyses were performed using SPSS version 21 (Chicago, IL) and MedCalc (MedCalc Software; Ostend, Belgium). In all instances, a P value of less than 0.05 was considered as the threshold for statistical significance.
The clinical characteristics of the 115 subjects are summarized in Table 1. Most were males (70%) and sex, age, and BMI were evenly distributed across subgroups, as projected by the recruitment design. Exercise performance evolved as expected with increasing COPD severity, with ventilatory limitation and gas exchange abnormalities becoming prominent in GOLD 3 and 4 patients.
Agreement in the determination of VT for human observers
There were no instances of “undetermined” VT.
Table 2 shows the agreement in the determination of V˙O2VT between human observers assessed using ICC. Overall, reliability between human observers was higher in control subjects than that in patients with COPD. In control subjects, ICC was 0.98 with V-slope and 0.98 with VEM, whereas in patients with COPD as a whole, ICC was 0.72 with V-slope and 0.64 with VEM. The 95% confidence intervals (CI) of ICC of the controls and patients with COPD were mutually exclusive. There was also a progressive decline in agreement between the two observers with increasing disease severity. In patients with GOLD 4 disease, agreement reached only “slight” levels.
ANOVA analysis revealed that the mean absolute differences in the measures of V˙O2VT using V-slope and VEM were statistically greater in patients with COPD compared with those in controls and that this difference increased with severity (Table 3).
Comparison of human and computer observers in the determination of V˙O2VT
E-Table 1 and e-Figure 1 (see Document, Supplemental Digital Content, Methods and Results, http://links.lww.com/MSS/A579) describe the results of the Passing–Bablok regression analysis comparing each human observer with the computerized analysis. In short, for both human observers, the relation of V˙O2VT with the computerized analysis did not differ from linearity, confirming that the data can be used in Passing–Bablok analysis. Using V-slope, V˙O2VT values from human observer 1 were interchangeable with computer analysis for controls but not for patients with COPD. Similar results were obtained using the VEM. In an identical manner, observer 2 was found to be interchangeable with computerized analysis when evaluating controls, but not when evaluating patients with COPD. Additional description of the Passing–Bablok regression analyses can be found in the supplemental digital content file (see Document, Supplemental Digital Content, Methods and Results, http://links.lww.com/MSS/A579).
Bland–Altman plots for both methods are shown in Figure 1 and similarly show that, although most data points remain inside the limits of agreement, patients with COPD generally have greater interobserver differences and wider dispersion of values than control subjects.
Interobserver differences in HRVT
Table 4 summarizes the interobserver differences in the evaluation of HRVT expressed both as absolute values and as a percentage of the peak HR attained during incremental exercise testing. Compared with controls using ANOVA, there was a statistically significant gradual increase in the interobserver difference of HRVT with disease severity. On average, only patients in the most severe COPD subgroup reached the prespecified threshold of clinical significance (±5 bpm). For each subgroup of patients, there were no significant differences in the mean interobserver difference in HRVT between patients with and without beta-blockers (see Document, Supplemental Digital Content, Methods and Results; Table 2, http://links.lww.com/MSS/A579).
Predictors of a larger interobserver difference in V˙O2VT
Table 5 describes the results of the stepwise multiple linear regression analysis. With V-slope, FEV1, % predicted and peak minute ventilation were the two sole independent predictors of a larger interobserver difference in V˙O2VT (R2 = 0.41), whereas with VEM, only FEV1 % predicted reached statistical significance (R2 = 0.50).
Intraobserver ICC measured on a subset of 50 patients showed relatively high reliability throughout the spectrum of disease severity (see Document, Supplemental Digital Content, Methods and Results; Table 3, http://links.lww.com/MSS/A579). For both observers and for both methods of observation, ICC across disease severity groups were all higher than 0.81.
To our knowledge, this is the first study to report on a direct evaluation of the reliability of human and computerized identification of V˙O2VT and HRVT across COPD severity groups. Our main results indicate that 1) reliability of human observers in the determination of V˙O2VT is lower in patients with COPD than that in controls, for both the V-slope and VEM methods, 2) human and computerized analyses of V˙O2VT are interchangeable in controls but not in patients with COPD, 3) FEV1 (percent predicted) is an independent predictor of a larger interobserver difference in measurement of V˙O2VT (with peak minute ventilation also being significant for V-slope), and 4) compared with that in controls, the increasing interobserver disparity in V˙O2VT assessment in patients with COPD corresponds to a gradually larger difference in the estimation of HRVT. These combined findings suggest that the baseline airflow obstruction and subsequent abnormalities in the ventilatory response of patients with COPD during exercise may be causative factors in the increasing variance of interobserver assessment of V˙O2VT (13,20). This supports a common impression among clinicians that, when represented graphically, the ventilatory parameters of patients with COPD produce more irregular and noisy patterns. Coupled with the fact that these patients show higher-than-predicted ventilation for any work rate, these anomalies seem to hinder the precise identification of a breaking point in the kinetics of ventilatory variables.
The available literature on this subject is scarce, especially in patients with COPD and has produced inconsistent results. Our results for control subjects are similar to those of Gladden et al. (17), who showed that in healthy volunteers, the intraobserver reliability of the VEM was high (ICC, 0.97), that the interobserver reliability (tested on nine observers) was lower (ICC, 0.70), and that agreement between a human observer and a computerized value of V˙O2VT was only moderate (ICC, 0.58). Filho et al. (12) also described similar results in a sample of 14 healthy subjects. In contrast, Garrard et al. (14) showed a higher intraobserver error when assessing V˙O2VT in healthy subjects. In this study, the interobserver error rates reached 29% and 24% using plots of the RER and V˙E but the V-slope and VEM performed better (19% and 15% error, respectively). Yeh et al. (41) described a mean range of 560 mL·min−1 among four observers trying to identify V˙O2VT using the VEM in healthy subjects. This is much larger than the difference found in our study in control subjects (44 mL·min−1).
Our results seem to be in line with those of Belman et al. (4) who studied the intra- and interobserver reliability of the determination of VT in patients with COPD using the V-slope method and the VEM on two separate exercise tests. They reported excellent intraobserver reliability for both method (Pearson correlation, 0.97 and 0.99 for the two observers) and good interobserver reliability for all methods (Pearson correlation, all higher than 0.74). However, their analysis was performed on a small uncharacterized subset (n = 14 at the maximum) of their overall cohort, which contained subjects with widely variable FEV1 values. In addition, the use of Pearson correlation to assess agreement between observers is often inappropriate (6). Our study used a larger, matched, well-characterized population and adds the findings of a progressive decline in interobserver reliability with disease progression and the poor relation between human and automated analysis in patients with COPD.
The clinical importance of the magnitude of interobserver differences identified can be put into perspective by comparing it with reported improvements in V˙O2VT after an exercise-training program. In patients with moderate to severe COPD, previous studies have documented improvements in V˙O2VT ranging from approximately 83 to 350 mL·min−1 after training (15,31,36,37). In our study, for moderate to very severe patients with COPD, interobserver differences in V˙O2VT ranged from 165 to 270 mL·min−1 using the V-slope and from 194 to 307 mL·min−1 using the VEM. It is therefore likely that interobserver differences in VT determination have an effect on the evaluation of changes in V˙O2VT after an exercise training program. In contrast, agreement for control subjects was much better (less than 50 mL·min−1 difference in V˙O2VT), suggesting that interobserver differences in VT play less role in this population (28). In addition, our findings concerning the low interobserver reliability of the determination of V˙O2VT should raise caution when using V˙O2VT or V˙E/V˙CO2 at VT as a prognostic marker in patients with heart failure or undergoing surgery if these subjects also have concomitant COPD.
Data concerning the reliability of computerized measurements of VT is limited. Most manufacturers of exercise testing equipment provide a unique software algorithm, and these different equations have been shown to provide variable estimates of VT both when using V-slope and the VEM (11). Any comparison of results originating from different software calculations must therefore be made with caution. Our data show that for control subjects, both human observers could be considered interchangeable with computer analysis, which is in line with the results of Santos et al. (30). In patients with COPD, however, human and automatic analyses were not interchangeable owing to significant systematic and proportional differences. This sheds an interesting light on the use of these computerized algorithms in daily practice, and clinicians may want to take it into account when assessing V˙O2VT using only automated reported values. Indeed, we believe these findings emphasize the need for clinicians to manually confirm any automated measurements of VT.
Our choice of using ±5 bpm as a threshold for a significant difference in HRVT was mostly empirical. It seems likely, however, that an error in measurement reaching 10 bpm would lead to important differences in the corresponding work rate or V˙O2. This estimation is difficult to quantify because the slope of the HR/V˙O2 relation during incremental exercise varies among individuals depending on baseline fitness level, use of negative chronotropic medication, or an underlying cardiopulmonary disease. A crude estimate of the effect of varying HR values on exercise intensity can be estimated using our cohort as a whole, where HRpeak was linearly related to peak work rate. Using this relation, a difference of 10 bpm in HR corresponded to an approximately 40-W difference in work rate, a difference that is arguably clinically significant, especially when considering patients with severe disease.
The optimal training intensity and modality for patients with COPD is an active matter of debate. Although current guidelines on pulmonary rehabilitation suggest using the American College of Sports Medicine framework for exercise prescription, they acknowledge that using standard “high-intensity” training may not be tolerable by patients with COPD (21) and that, in this context, a training program based on perceived exhaustion (Borg scale rating 4–6) is adequate (33). Coincidentally, this level of perceived exhaustion is known to correspond to V˙O2VT (43). Moreover, the safety and efficacy of using fixed percentages of HR or V˙O2peak as training targets have been challenged by recent publications (10,19). Our data show that the absolute interobserver difference in HRVT in patients with COPD becomes increasingly large as disease worsens when compared with that in controls. In this context, if HRVT was used a marker for exercise intensity prescription in patients with very severe COPD, this target could translate into an unacceptably large array of actual training intensity, which could respectively result in exercise inducing undue fatigue and intolerance (above VT) or of too low intensity to provide benefits (below VT). In the other subgroups of patients, the interobserver difference in HRVT was less important and is therefore less likely to negatively affect a training regimen.
This study has several limitations. First, VT measurement was performed on retrospectively collected data, and as such, there is a chance of selection bias. We tried to limit this effect by matching subjects on several relevant clinical parameters. Second, the external validity of the study is impaired by our choice of including only patients having performed at least 6 min on incremental exercise testing, a duration which might not be routinely sustained by the patients with the most severe COPD. We believe this criterion is valid in a proof-of-concept framework, with the goal of maximizing the attainment of a true VT, but needs to be taken into account when applying these results to a wider population with COPD. Third, we recognize that the relative lack of experience of the observers (<5 yr) may be of concern. Hansen et al. (18) showed that when measuring VT in patients with pulmonary hypertension, the agreement between two experienced observers was better than the one between inexperienced observers. However, the overall difference in agreement was small. Whereas experienced observers had a mean difference of 20 mL·min−1 in their measure of V˙O2VT between them, less experienced observers had a mean difference of 60 mL·min−1. The clinical relevance of such a small difference is unclear. The fact that our two observers strongly agreed with each other in patients with milder disease maintained relatively high intraobserver reliability is reassuring.
Fifth, the choice of using the V-slope and the VEM to measure VT was based on the abundance of their use in the literature and the available data regarding their reliability in healthy subjects. Current guidelines suggest the use of either techniques when measuring VT (1). Although other methods to assess VT have been reported, they are often lacking a standardized definition, scarcely used or known to relate closely to the V-slope or the VEM (i.e., changes in PETO2 and PETCO2 vs time, V˙E, V˙O2, V˙CO2 or RER vs work rate, V˙E vs V˙CO2, HR inflection point) (5). Therefore, we believe that the choice of using these two methods is representative of common clinical practice and allows a more thorough comparison the available literature. Sixth, our choice of reporting data using 20-s averages of breath-by-breath data could be criticized. This parameter was chosen in accordance with current guidelines concerning the reporting of data during CPET (1), as it allows, in our opinion, a balance between “noise” and the clear representation of respiratory kinetics. Whether the use of different time-averaging intervals could further influence the detection of VT requires further studies. Finally, the differences in ramp increment rate between subgroups are an expected finding, and whether this may have had an effect on the determination of V˙O2VT is unclear. However, studies have reported the lack of significant differences in the determination of V˙O2VT between ramp increments of 7–23 W·min−1 in patients with heart failure (2) and between increments of 20–50 W·min−1 in young healthy subjects (9).
In conclusion, results from the present study show that the agreement between human observers in the determination of VT in patients with COPD is lower than that in controls, that human and computer analyses of V˙O2VT are not interchangeable in these patients, and that these findings are directly related to the severity of airflow obstruction. Furthermore, the decline in precision in the identification of V˙O2VT corresponds to an increasing variability when evaluating HRVT. Clinicians should be aware of the discrepancy between software and human identification of VT when reporting automated values of V˙O2VT, and these findings should be taken into account when using VT for exercise prescription, as a tool to monitor response to an intervention, or as a prognostic marker in patients with COPD.
Dr. Pepin was supported by a grant from the Fonds de Recherche du Québec–Santé.
The authors have no conflict of interest to declare.
The results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. ATS/ACCP statement on cardiopulmonary exercise testing. Am J Respir Crit Care Med
. 2003; 117: 211–77.
2. Agostoni P, Bianchi M, Moraschi A, et al Work-rate affects cardiopulmonary exercise test results in heart failure. Eur J Heart Fail
. 2005; 7( 4): 498–504.
3. Beaver WL, Wasserman K, Whipp BJ. A new method for detecting anaerobic threshold by gas exchange. J Appl Physiol (1985)
. 1986; 60( 6): 2020–7.
4. Belman MJ, Epstein LJ, Doornbos D, Elashoff JD, Koerner SK, Mohsenifar Z. Noninvasive determinations of the anaerobic threshold. Reliability and validity in patients with COPD. Chest
. 1992; 102: 1028–34.
5. Binder RK, Wonisch M, Corra U, et al Methodological approach to the first and second lactate threshold in incremental cardiopulmonary exercise testing. Eur J Cardiovasc Prev Rehabil
. 2008; 15: 726–34.
6. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet
. 1986; 1( 8476): 307–10.
7. Borg GA. Psychophysical bases of perceived exertion. Med Sci Sports Exerc
. 1982; 14( 5): 377–81.
8. Casaburi R, Patessio A, Ioli F, Zanaboni S, Donner CF, Wasserman K. Reductions in exercise lactic acidosis and ventilation as a result of exercise training in patients with obstructive lung disease. Am Rev Respir Dis
. 1991; 143( 1): 9–18.
9. Davis JA, Whipp BJ, Lamarra N, Huntsman DJ, Frank MH, Wasserman K. Effect of ramp slope on determination of aerobic parameters from the ramp exercise test. Med Sci Sports Exerc
. 1982; 14( 5): 339–43.
10. Díaz-Buschmann I, Jaureguizar KV, Calero MJ, Aquino RS. Programming exercise intensity in patients on beta-blocker treatment: the importance of choosing an appropriate method. Eur J Prev Cardiol
. 2014; 21( 12): 1474–80.
11. Ekkekakis P, Lind E, Hall EE, Petruzzello SJ. Do regression-based computer algorithms for determining the ventilatory threshold agree? J Sports Sci
. 2008; 26( 9): 967–76.
12. Filho P, Pompeu F. Silva APRdSe. Accuracy of VO2max
and anaerobic threshold determination. Rev Bras Med Esporte
. 2005; 11( 4): 162–5.
13. Frisk B, Espehaug B, Hardie JA, et al Airway obstruction, dynamic hyperinflation, and breathing pattern during incremental exercise in COPD patients. Physiol Rep
. 2014; 2( 2): e00222.
14. Garrard C, Das R. Sources of error and variability in the determination of anaerobic threshold in healthy humans. Respiration
. 1987; 51: 137–45.
15. Gimenez M, Servera E, Vergara P, Bach JR, Polu JM. Endurance training in patients with chronic obstructive pulmonary disease: a comparison of high versus moderate intensity. Arch Phys Med Rehabil
. 2000; 81: 102–9.
16. Gitt AK, Wasserman K, Kilkowski C. Exercise anaerobic threshold and ventilatory efficiency identify heart failure patients for high risk of early death. Circulation
. 2002; 106: 3079–84.
17. Gladden LB, Yates JW, Stremel RW, Stamford BA. Gas exchange and lactate anaerobic thresholds: inter- and intraevaluator agreement. J Appl Physiol (1985)
. 1985; 58( 6): 2082–9.
18. Hansen JE, Sun XG, Yasunobu Y, et al Reproducibility of cardiopulmonary exercise measurements in patients with pulmonary arterial hypertension. Chest
. 2004; 126: 816–24.
19. Hofmann P, Tschakert G. Special needs to prescribe exercise intensity for scientific studies. Cardiol Res Pract
. 2010; 2011: 209302. doi: 10.4061/2011/209302.
20. Loring SH, Garcia-Jacques M, Malhotra A. Pulmonary characteristics in COPD and mechanisms of increased work of breathing. J Appl Physiol (1985)
. 2009; 107: 309–14.
21. Maltais F, Leblanc P, Jobin J, et al Intensity of training and physiologic adaptation in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med
. 1997; 155: 555–61.
22. Myers J, Gullestad L, Vagelos R, et al Clinical, hemodynamic, and cardiopulmonary exercise test determinants of survival in patients referred for evaluation of heart failure. Ann Intern Med
. 1998; 129: 286–93.
23. Nery LE, Wasserman K, Huntsman DJ, Andrews JD, Hansen JE, Whipp BJ. Ventilatory and gas exchange kinetics during exercise in chronic airways obstruction. J Appl Physiol Respir Environ Exerc Physiol
. 1982; 53( 6): 1594–602.
24. Ozcelik O, Ward SA, Whipp BJ. Effect of altered body CO2
stores on pulmonary gas exchange dynamics during incremental exercise in humans. Exp Physiol
. 1999; 84( 5): 999–1011.
25. Passing H, Bablok. A new biometrical procedure for testing the equality of measurements from two different analytical methods. Application of linear regression procedures for method comparison studies in clinical chemistry, Part I. J Clin Chem Clin Biochem
. 1983; 21( 11): 709–20.
26. Péronnet F, Thibault G, Rhodes EC, McKenzie DC. Correlation between ventilatory threshold and endurance capability in marathon runners. Med Sci Sports Exerc
. 1987; 19( 6): 610–5.
27. Piepoli MF. Exercise tolerance measurements in pulmonary vascular diseases and chronic heart failure. Respiration
. 2009; 77: 241–51.
28. Poole DC, Gaesser GA. Response of ventilatory and lactate thresholds to continuous and interval training. J Appl Physiol (1985)
. 1985; 58 (4): 1115–21.
29. Reinhard U, Müller PH, Schmülling RM. Determination of anaerobic threshold by the ventilation equivalent in normal individuals. Respiration
. 1979; 38: 36–42.
30. Santos EL, Giannella-Neto A. Comparison of computerized methods for detecting the ventilatory thresholds. Eur J Appl Physiol
. 2004; 93( 3): 315–24.
31. Serres I, Varray A, Vallet G, Micallef JP, Préfaut C. Improved skeletal muscle performance after individualized exercise training in patients with chronic obstructive pulmonary disease. J Cardiopulm Rehabil
. 1997; 17( 4): 232–8.
32. Shrout P. Measurement reliability and agreement in psychiatry. Stat Methods Med Res
. 1998; 7: 301–17.
33. Spruit MA, Singh SJ, Garvey C, et al An official American Thoracic Society/European Respiratory Society statement: key concepts and advances in pulmonary rehabilitation. Am J Respir Crit Care Med
. 2013; 188( 8): e13–e64.
34. Svedahl K, MacIntosh BR. Anaerobic threshold: the concept and methods of measurement. Can J Appl Physiol
. 2003; 28( 2): 299–323.
35. Torchio R, Gulotta C, Parvis M, et al Gas exchange threshold as a predictor of severe postoperative complications after lung resection in mild-to-moderate chronic obstructive pulmonary disease. Monaldi Arch Chest Dis
. 1998; 53( 2): 127–33.
36. Vallet G, Ahmaïdi S, Serres I, et al Comparison of two training programmes in chronic airway limitation patients: standardized versus individualized protocols. Eur Respir J
. 1997; 10: 114–22.
37. Vallet G, Varray A, Fontaine JL, Préfaut C. Value of individualized rehabilitation at the ventilatory threshold level in moderately severe chronic obstructive pulmonary disease. Rev Mal Respir
. 1994; 11( 5): 493–501.
38. Wasserman K. Determinants and detection of anaerobic threshold and consequences of exercise above it. Circulation
. 1987; 76 ( VI Suppl Pt 2): VI29–39.
39. Wasserman K, Whipp BJ, Koyl SN, Beaver WL. Anaerobic threshold and respiratory gas exchange during exercise. J Appl Physiol
. 1973; 35( 2): 236–43.
40. West MA, Lythgoe D, Barben CP, et al Cardiopulmonary exercise variables are associated with postoperative morbidity after major colonic surgery: a prospective blinded observational study. Br J Anaesth
. 2014: 112( 4) 665–71.
41. Yeh MP, Gardner RM, Adams TD, Yanowitz FG, Crapo RO. “Anaerobic threshold”: problems of determination and validation. J Appl Physiol Respir Environ Exerc Physiol
. 1983; 55( 4): 1178–86.
42. Zacarias EC, Neder JA, Cendom SP, Nery LE, Jardim JR. Heart rat the estimated lactate threshold in patients with chronic obstructive pulmonary disease: effects on the target intensity for dynamic exercise training. J Cardiopulm Rehabil
. 2000; 20( 6): 369–76.
43. Zamunér AR, Moreno MA, Camargo TM, et al Assessment of subjective perceived exertion at the anaerobic threshold with the Borg CR-10 scale. J Sports Sci Med
. 2011; 10: 130–6.