To the Editor:
Pulmonary arterial hypertension (PAH) remains a disabling and frequently lethal disease despite remarkable advances in treatment. Cardiopulmonary exercise testing (CPET) has proved a valuable tool to objectively quantify disease severity and estimate prognosis in these patients [1–3].
Exercise intolerance is characteristically multifactorial in PAH. Among its potential contributing mechanisms, increased ventilatory response, deranged pulmonary mechanics, peripheral muscle impairment and reduced oxygen delivery have been more widely investigated [1–5]. In this context, a CPET-derived variable that conflates the effects of increased ventilation and poor O2 transfer and/or peripheral O2 utilisation is the O2 uptake efficiency slope (OUES) [6]. OUES is the slope of the linear relationship between O2 uptake (V′O2) and the logarithmic transformation of minute ventilation (V′E) during rapidly incremental exercise, i.e. it aims to reflect how effectively O2 is extracted from the atmosphere and taken into the body as exercise progresses. We recently found that a combination of increased sub-maximal exercise V′E as a function of carbon dioxide output (V′CO2) and reduced O2 delivery/utilisation (as suggested by shallow V′O2–work rate relationship) were independent predictors of negative outcome in PAH of mixed aetiology [7]. These findings prompted the hypothesis that OUES would combine the prognostic information provided separately by those variables, thereby being the single predictor of poor outcome in our cohort.
In order to address this question, we revisited our dataset and contrasted OUES prognostic relevance with that of a range of resting and cardiopulmonary exercise responses to ramp-incremental cycle ergometry. In the previous report [7], we described results from a group of 84 patients in whom 16 PAH-related deaths and two atrial septostomies were observed in a 5-year follow-up. In this present article, we extend these observations to 98 patients (70 females, 48 with idiopathic PAH and 50 with associated PAH) followed for up to 6.5 years, in whom 17 PAH-related deaths and four atrial septostomies were recorded. Receiver operating characteristic (ROC) curves were used to obtain the best cut-offs for prognostication. Cox proportional hazards and Kaplan–Meier cumulative survival analyses were performed following standard procedures. The statistical significance for all tests was p<0.05.
We found that event-positive patients showed lower OUES than their counterparts (mean±sd 0.72±0.27 L·min−1 per logV′E versus 0.91±0.31 L·min−1 per logV′E, p<0.05). OUES was significantly related to prognosis in the univariate analysis (hazard ratio (HR) (95% CI) 9.5 (3.8–23.8), p<0.01). The ROC curve analysis (area under the curve 0.688, 95% CI 0.542–0.833; p<0.01) indicated a best OUES cut-off for prognostication of 0.56 L·min−1 per logV′E. Among a range of resting and exercise variables, only peak O2 uptake (V′O2peak), V′E/V′O2peak, change in (Δ)V′E/ΔV′CO2 and ΔV′O2/Δwork rate were also predictors of poor outcome in the univariate analysis (HR (95% CI) 2.4 (1.1–5.7), 8.4 (1.2–55), 4.2 (1.5–11.6) and 6.8 (2.6–17.3), respectively; p<0.05). Considering the close similarity between OUES and V′E/V′O2peak, multivariate regression analyses were performed using either of these variables. The prognostic power of OUES (HR (95% CI) 4.63 (1.38–15.5), p=0.01) overcame that of V′O2peak and the slopes of ΔV′E/ΔV′CO2 and ΔV′O2/Δwork rate in the final model (HR (95% CI) 1.74 (0.69–4.37), p=0.23; 2.20 (0.73–6.67), p=0.16; and 2.24 (0.70–7.11), p=0.17; respectively). ΔV′O2/Δwork rate, however, remained in the final model when V′E/V′O2peak was considered instead of OUES (p<0.05). In line with our hypothesis, therefore, OUES (but not V′E/V′O2peak) conflated the prognostic information provided by ΔV′E/ΔV′CO2 and ΔV′O2/Δwork rate in our previous study (fig. 1). The same results were found when only mortality was considered the main outcome (data not shown).
Some preliminary considerations about the ventilatory–metabolic coupling during incremental exercise might be instructive to interpret our results. The ventilatory response to exercise is tightly coupled with the rate at which CO2 is exchanged at the lungs, i.e. V′CO2. V′E is needed to clear a given amount of CO2 when its arterial partial pressure is regulated at a lower set-point and the dead space fraction of the breath is increased, as found in PAH [1–5]. Progressive lactic acidosis (i.e. beyond the point at which the bicarbonate reserve is sufficient to fully tamponade H+) brings additional source of V′E stimuli at abnormally low levels of exertion in these patients. Right-to-left shunt through an open foramen ovale might also expose the carotid bodies to an extra burden of H+, which might sum up with variable levels of hypoxaemia. In fact, 21 (21%) out of 98 patients showed evidence of an open foramen ovale during exercise. As expected, ΔV′E/ΔV′CO2 was greater and OUES lower in these patients compared with their counterparts (82±25 versus 52±12 and 0.65±0.16 L·min−1 per log V′E versus 0.90 0.29 L·min−1 per log V′E, respectively). However, foramen ovale opening was not predictive of negative outcome in our sample. Additional sources of increasing nonmetabolic drive include excessive sympathetic drive, respiratory muscle weakness, ergoreceptor stimuli and increased pulmonary artery/right ventricle pressures [1–5]. These premises fully justify the notion that excessive V′E response to CO2 (ΔV′E/ΔV′CO2 slope) is a valuable marker of disease severity and progression in PAH [1–5, 7].
OUES is substantially more complex to interpret than ΔV′E/ΔV′CO2. Therefore, while V′E is appropriately considered the dependent variable in ΔV′E/ΔV′CO2, it is the independent parameter in OUES. Consequently, part of the information provided by OUES is intrinsically linked to the mechanisms regulating CO2 exchange. However, the rate at which O2 is taken into the body during the incremental phase of exercise (the dependent variable in OUES) is not limited by the ventilatory response in PAH but rather dependent upon cardiovascular, haematological and muscular adjustments [1–5]. In other words, abnormalities in any of the determinants of exercise hyperpnoea during exercise (either metabolic or nonmetabolic) plus derangements in systemic O2 transfer might reduce OUES. The inordinate high ventilatory response in PAH coupled with severe reductions O2 delivery/utilisation might explain why OUES seems to have a higher prognostic relevance in these patients than in chronic heart failure [8]. In fact, whereas OUES cut-offs to indicate poor prognosis in chronic heart failure were in the range of 1.3–1.5 L·min−1 per logV′E [9], substantially lower values should be used in PAH (0.56 L·min−1 per logV′E).
OUES calculation in PAH, however, may have some caveats that should be carefully considered. It should be noted that log transformation works particularly well in pronouncedly skewed data. If this is not the case, it may overcompensate a right-skewed data set and create a left-skewed one [10]. In practice, these caveats might promote nonlinearities and artificially increase OUES. This would be particularly anticipated in patients with high exercise V′E from the start of the test who are able to sustain exercise long above the “anaerobic” threshold. In fact, Niemeijer et al. [11] found that the anaerobic threshold impacted on OUES linearity in chronic heart failure. Moreover, Williamson et al. [12] reported that OUES increased as exercise progressed, with a peak value of V′CO2/V′O2 of ∼1. The best approach to calculate OUES in PAH patients with extremely elevated V′E responses during exercise tests that are not too short (i.e. 8–12 min of incremental phase) remains to be determined. V′E/V′O2peak is a more straightforward variable and it might provide an interesting alternative to OUES. As a discrete, end-exercise variable, however, V′E/V′O2peak is expected to be more influenced by effort (or early exercise cessation due to intolerable dyspnoea) and degree of ventilatory response to lactic acidosis than OUES. Additional studies aiming to contrast the prognostic relevance of OUES and V′E/V′O2peak are therefore warranted.
In conclusion, OUES is a powerful prognostic index in patients with PAH. Although this variable is not usually available during CPET, most of the commercially available systems allow post-test data transformation making its calculation clinically friendly. Future investigations, however, should address whether restraining OUES estimation to specific time frames (e.g. below or above any inflection point) would further increase its prognostic relevance in PAH.
Footnotes
Conflict of interest: None declared.
- Received September 25, 2013.
- Accepted November 5, 2013.
- ©ERS 2014