Controlled versus free breathing for multiple breath nitrogen washout in healthy adults

Summary of results

In this study in healthy adults comparing two commonly used MBNW breathing protocols, we found that: 1) FRC was comparable between the two protocols, however indices of ventilation heterogeneity based on phase III slopes (S_III) were not – S_cond was poorly correlated between the two protocols and S_acin was systematically higher under the free breathing protocol, whereas LCI was only marginally higher; 2) there was greater between-protocol discrepancy in S_acin in subjects whose free-breathing pattern deviated from that of the volume- and frequency-controlled protocol; and 3) S_cond and S_acin exhibited poorer between-session repeatability under the free breathing protocol.

Lack of discrepancy in FRC

Yammine et al. [12] demonstrated in children that S_cond and LCI obtained from a controlled breathing protocol were higher than when obtained by free breathing and that the discrepancy was related to a decrease in FRC during controlled breathing. Thus, the authors argued that a controlled breathing protocol may overestimate heterogeneity by inadvertently causing individuals to change the EELV at which they would otherwise normally breathe. In children, a 1 L V_T represents a significant proportion of total lung capacity; the high V_T may increase ventilation heterogeneity due to the recruitment of normally non-ventilated lung units [19], or due to closure of lung units in the dependent regions if the child expires to below their normal EELV [20], or some combination of both.

Our current findings in healthy adults do not support the argument that FRC is altered with controlled breathing, as we showed that FRC estimated by the two protocols was in fact comparable, with no systematic bias. Furthermore, both were comparable to the “gold standard”, i.e. FRC_pleth. A number of factors may explain the differences between our findings and those of Yammine et al. For example, the effect of variation in FRC may be relatively mild in the adults in our study compared to that in children, and thus insufficient to cause significant differences in heterogeneity. We observed that the variability in between-protocol differences (assessed by the LOA) in FRC were comparable to the between-session variabilities of either protocol, though interestingly between-session LOA was numerically higher under the controlled breathing protocol. Alternatively, our standard procedure for the controlled breathing protocol includes an initial period of unconstrained breathing to allow the individual's EELV to stabilise before commencing the washout. This is in line with current ATS/ERS recommendations [3] and may have ensured FRC differences were kept minimal between protocols.

Discrepancy in ventilation heterogeneity indices

While LCI – an index of global heterogeneity that is independent of S_III – was comparable between protocols (with a small bias), we found that S_III-dependent ventilation heterogeneity partitioned into proximal/conducting (S_cond) and distal/acinar (S_acin) zones was not. S_cond was poorly correlated between protocols, perhaps driven by a large between-session variability. S_acin was significantly different between protocols; moreover, S_acin obtained by free breathing was higher than that obtained by controlled breathing in individuals with relatively small V_T or high RR. Between-protocol differences in S_acin appear to be independent of age, height or lung size, i.e. a change in breathing pattern was the dominant driver of the discrepancy.

There are two possible explanations for this. First, the effect could be purely methodological: when breaths are too “shallow”, estimation of S_III is rendered invalid as the N₂ expirogram either never reaches a plateau or attains a plateau that is too short for reliable curve fitting, which would subsequently affect the derivation of S_acin and S_cond. Anecdotal evidence suggests this occurs frequently. Methods to more reliably estimate S_III exist [21] but are still ultimately dependent on the presence of distinct phase II and III regions in the expirogram and an adequate portion of phase III. More useful may be methods of quantifying ventilation heterogeneity that do not rely on S_III estimation [22, 23] – these may help not just in free breathing but also in pathologies where S_III is often difficult to define. The feasibility and clinical relevance of these methods should be a direction for future investigation.

Alternatively, the effect may be physiological: when patients breathe shallower and faster, the fast-emptying lung compartments will increase their N₂ washout while the slow-emptying lung compartments will be unable to empty completely, thus increasing apparent heterogeneity estimated by S_acin and S_cond. Indeed, the foundational studies on ventilation distribution clearly demonstrated an effect of breathing pattern (including V_T) on MBNW indices [4]. Recently, Ratjen and colleagues [24] investigated the effects of altered V_T on MBNW indices and found that shallow breathing (compared to “unrestricted” breathing) significantly increased LCI. The authors speculated that this was due to the effects of increased dead space to V_T ratio on the determinants of LCI (both CEV and FRC). Notably, these investigators did not examine the impact on S_cond or S_acin.

We found that, on average, V_T was shallower, RR was faster and CEV higher in the free breathing protocol in our study participants. The change in CEV would be expected, since smaller breaths may reduce alveolar gas mixing efficiency and therefore increase the volume (and time, though this was not significant) required to reach end-of-test criteria. This likely drove the changes in LCI we observed – however, differences in LCI were unrelated to differences in CEV (supplementary figure S5). It should be noted that the differences seen in LCI were very small relative to its magnitude and unlikely to be clinically significant. Furthermore, we found a weak relationship between differences in S_cond and differences in CEV; this was not seen with S_acin (supplementary figure S6). However, when plotted against differences in CEV/FRC (i.e. LCI, which corrects CEV for lung volume), the relationship with S_acin became significant. This supports the interpretation that the differences seen in S_acin were driven by changes in ventilation heterogeneity, rather than alterations to lung volume.

Variability of MBNW indices

We also report between-session variability over 2–10 weeks for both protocols, using the same commercially available device. This allowed us to compare between-protocol differences in MBNW indices against their short-term variability. Within-session CoV of FRC and LCI was similar to published values for the free breathing protocol [25], using the same device. However the between-session coefficient of repeatability (equal to 1.96×sd of the mean differences) of LCI, S_cond and S_acin was greater in our study, which may reflect the different time between sessions (weeks/months versus days). There is a paucity of repeatability data in health, and previous reports may not be generalisable due to persistent between-device differences [17, 18, 26]. Nevertheless, the excellent between-session repeatability for FRC and LCI [25, 27–29] but poor repeatability in S_acin and S_cond [25, 28] has been noted with the free breathing protocol by other investigators, regardless of device, tracer gas or disease. The higher between-session repeatability values seen in S_cond and S_acin with the controlled breathing protocol are consistent with other studies [30, 31].

It is also worth noting that FRC and LCI are derived from successive cumulation of volume at each breath, i.e. they are integrated measures, which are more robust to noise. On the other hand, S_cond and S_acin are estimated from slopes, i.e. differentiated measures, which are inherently susceptible to noise. Furthermore, unlike FRC and LCI, the values of S_cond and S_acin are very close to zero, which may explain the very low between-session ICC values observed. These may be fundamental reasons for the greater variability seen in S_cond and S_acin but not FRC and LCI, and this variability becomes even more pronounced when variability in tidal breathing is introduced. This higher variability may also drive the differences seen between the two protocols. We also note the potential for these differences to be further exaggerated in disease – this and the high between-session variability observed may limit the clinical utility of S_III indices derived from free breathing. These speculations need to be confirmed in further investigations.

Significance

Both MBNW breathing protocols are designed to measure ventilation heterogeneity during resting tidal breathing. The controlled breathing protocol was originally devised based on modelling studies in which convection-dependent and diffusion-convection interaction-dependent mechanisms of airflow in the lung (the basis of S_cond and S_acin, respectively) were first described [32]. The controlled breathing protocol standardises the V_T at which ventilation distribution is assessed, and also ensures that expiration occurs well into the alveolar plateau so that a reliable SIII estimation can be made. This standardisation potentially minimises variability between tests. However, this “artificial” pattern of breathing may distribute the ventilation differently in lungs of different sizes, i.e. in shorter versus taller people. These differences may then affect the measurements of MBNW indices such that they no longer reflect true resting or “natural” ventilation distribution. In contrast, under a free breathing protocol, breathing occurs at (or closer to) natural resting breathing pattern; thus, the MBNW indices obtained arguably better reflect the individual's native ventilation distribution during normal tidal breathing.

Our results lend weight to the idea that controlled breathing imposes a condition that is different to the individual's natural resting breathing pattern, and that MBNW indices derived from this protocol may not necessarily reflect their “native” ventilation heterogeneity. Specifically, breathing at 1 L tended to improve heterogeneity measured by S_acin in individuals whose V_T during free breathing was <1 L, or whose RR during free breathing was faster than the prescribed 8–12 breaths per minute. So, which is the “better” test? We suggest that it is whichever provides the highest sensitivity in detecting disease or assessing treatment. The “better” test may even be disease-specific and warrants further investigation.

Since these differences are apparent despite the use of V_T-corrected versions of S_cond and S_acin, our study suggests that the linear V_T correction currently employed for the free breathing protocol may not be adequate. Again, methods which assess ventilation heterogeneity without the need for estimating S_III, ideally in a manner that is independent of V_T [22], may be useful here.

Limitations

Our study has several limitations to be taken into consideration. First, a number of the volunteers could be considered “trained” in lung function testing. However, it is unlikely that prior experience would influence the largely effort-independent MBNW test. Second, 4 out of 27 participants required coaching before or after free breathing testing due to insufficient expired volume for S_III estimation. Thus, there is a chance our results may actually underestimate the true discrepancy between the two protocols, particularly in shallow breathers [24] or in lung disease where increased dead space further complicates S_III estimation [33]. Third, our results only apply to a healthy population tested on a single MBNW device. There are likely to be differences between equipment [17, 18] and between populations (e.g. children, older adults, lung disease) that limit the generalisability of our results.

Conclusion

We have demonstrated that while controlled and free breathing protocols are equivalent in terms of FRC and LCI, phase III-derived MBNW indices of ventilation heterogeneity are lower (i.e. better) when measured by the controlled breathing method. This effect seems to be driven largely by deviations in breathing pattern between the two protocols, particularly in patients who either breath faster or shallower than the criteria imposed by the controlled breathing protocol. Future work needs to be carried out in diseased patients to see if these observations hold true. Our study sheds light on potential physiological mechanisms behind these differences, and the overall interpretation of ventilation heterogeneity measured by MBNW. In better characterising the difference between the two protocols, these findings also help facilitate the ongoing efforts to standardise MBNW as an emerging clinical test.