Reproducibility of cardiopulmonary exercise test variables: getting into an additional strength of the test

Exercise is the primary source of symptom exacerbation, and intolerance to physical activity is one of the most disabling problems experienced by patients with cardiopulmonary disorders. Exercise evaluation combined with gas exchange analysis by cardiopulmonary exercise testing (CPX) can assist the clinician with a wealth of information beyond standard exercise electrocardiography testing.1 This approach is useful in terms of assessing the mechanism of exercise intolerance, evaluating disability, making activity and training recommendations, quantifying responses to therapy, and predicting outcome. For these distinctive features, CPX has received a particular amount of attention in recent years; numerous studies have adopted CPX variables as endpoints over the last decade and the testing, in all its features, is becoming more common in daily clinician practice.1 Given the key role of exercise impairment in estimating cardiac and non-cardiac risk,2 objective and reproducible methods for exercise capacity assessment are needed. Therefore, to be considered as a reliable tool in the follow-up assessment of cardiopulmonary diseases the performance of CPX variables must be proven to be reproducible across different cohorts of patients. In the present issue of the Journal, Barron et al.3 contribute to our understanding on the repeatability of CPX data by reporting test to test evaluation in a series of patients with cardiac (heart failure, mitral valve disease) and chronic obstructive pulmonary disease (COPD). The study builds on prior evidence of very good test reproducibility4–8 confirming that mainstay CPX measurements, such as oxygen consumption (VO2) and ventilation to carbon dioxide production (VE/VCO2), are well-reproducible and extends these findings with some remarkable new evidence. Investigators report on the largest number of patients (93) ever tested who underwent two CPX tests, the first as familiarisation at a fixed ramp protocol (10 watt/min) and the second, performed two hours later or sometimes on a separate day, using a protocol to achieve the maximum in 8–10 min duration based on the results of the first test. The analysis incorporated a large set, practically all, of gas exchange variables directly measured or derived and the statistical analysis was performed using robust procedures by testing the interclass correlation coefficient (ICC) and the coefficient of variation between tests. Of note, this is the first report assessing test to test repeatability of some variables that are derived and highly prognostic especially in heart failure patients, e.g. oxygen uptake efficiency slope (OUES),9 circulatory power10 and exercise oscillatory ventilation (EOV).11 In addition, the repeatability of ventilation efficiency was determined by testing the three methods proposed in the literature, VE vs VCO2 slope, their nadir ratio and the ratio at the anaerobic threshold (AT). Interestingly, the reproducibility end-point was not limited to variable comparison per se, but extended to ascertain the role of protocol, disease aetiology and inter-test interval. Also, the potential modifier role of other factors, e.g. age, body mass index (BMI) and gender, was taken into account. Peak VO2, O2 pulse and the VE/VCO2 ratio at nadir showed the best repeatability performance, with within-subject coefficient of variation <0.12. Good repeatability was maintained for most other variables without any significant influence of age, gender, BMI and inter-test interval, with the exception of EOV and VO2/work rate (WR) in older patients. Some specific considerations on these variables seem worthy of note since both are calculated rather than measured and may be operator-dependent and affected by the methodology. EOV was not observed in the COPD patient cohort, whereas it occurred in 23 (33%) cardiac patients. In eight subjects of this subgroup the EOV criteria were not reached at test 2. The adopted criterion for EOV calculation was the one proposed by Leite and coworkers12 that is not the method suggested by current statements.1,13 Nevertheless, a comparative analysis with other proposed criteria would have been useful to understand the true reproducibility of a meaningful aberrant phenomenon that strongly predicts an adverse outcome.11,14 On the other hand, since EOV depends on a complex interaction between a delayed circulatory transit time and abnormal neural ventilatory control, some biological variability may be part of the random error affecting reproducibility, thus requiring more than one CPX test evaluation. VO2/WR is a meaningful physiological variable that reflects the amount of aerobically generated adenosine triphosphate for the given work. The correct modality for measuring VO2/WR would require beginning calculation 1 min after WR started to increase. The one-minute delay after the start of increase in WR is used to take into account the time constant for VO2 to respond to the increasing WR (∼35 s for normal subjects).15 The criteria used in the present study for measuring VO2/WR are not specified. In patients with severe cardiac systolic dysfunction and mitral insufficiency, it is quite frequent to observe a flattening in the expected linear relationship.16 Numbers of patients in which a flattening pattern was observed and when it occurred are missing information that appear useful for a thorough definition of how much VO2/WR is reproducible. A further caveat is the repeatability of VO2 at AT that may be be quite unreliable in the presence of exercise-induced right to left shunt and O2 desaturation. A typical example is the patient with severe pulmonary arterial hypertension.7 In the COPD population, VO2 at AT was reproducible and no significant O2 desaturation was observed which is suggestive of a mild severity of lung disease. Caution is, however, advised when applying these findings to patients with severe COPD and significant shunting during exercise. Overall, these positive and comprehensive results provided by Barron et al. likely would have given even better test to test repeatability if a hypothetical test 3 could have been performed, making an inter-test comparison between test 2 and 3 and excluding the familiarisation one. Actually, test 1 vs 2 performed much better compared to a previous study17 and this has been explained with the use of bicycle rather than the treadmill, where familiarisation may be less important. Nonetheless, familiarisation with a CPX test means not only mode of exercise but also adaptation to the mouthpiece or mask. So what are the implications to this study of the use of CPX by the clinical researcher or clinician? Clearly, present findings provide a step forward in consolidating the reliability of the technique. The main message is that clinical trials that utilise CPX variables as outcomes do not have to account for a significant learning effect compared to other types of tests, for example six minute walk test,18,19 when designing their study protocols and interpreting results. In clinical scenarios where it is crucial to discern viable candidates for advanced and costly surgical therapies (e.g. left ventricular assist devices and heart transplantation) a single CPX evaluation should be considered a valuable addition in the clinical decision-making process. Future research should focus on other conditions and relevant confounding factors especially addressing whether, in the presence of growing epidemic disorders such as heart failure preserved ejection fraction and comorbid states associated with a high rate of cardiovascular events (primarily diabetes, obesity and atrial fibrillation), gas exchange analysis still provides a high level of test to test concordance. Until then, clinicians need to be aware that reproducibility is an additional strength of CPX enabling diagnostic and therapeutic decisions. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. None declared.