Reference equations for pulmonary diffusing capacity of carbon monoxide and nitric oxide in adult Caucasians

Comparison to other reference equations

As seen in figure 2, the reference equations of the present study estimate comparable, but for D_LNO somewhat lower values, to those obtained if using one of the other reference equations for adults [12–15, 31]. As described in the Methods section, quality control of measurements was an integrated part of this study and both sets of equipment passed all tests performed. In addition, the fact that no statistically significant difference was found when comparing measurements from the two sets of equipment strengthens our belief that both measured correctly throughout the study.

Of the five reference equations compared in figure 2, Aguilaniu et al. [14] produced the highest predicted values. Several circumstances might work together to explain the difference between their results and the results from the present study. Firstly, Aguilaniu et al. [14] performed their measurements at two different sites (the cities of Grenoble and Bordeaux) and reported significant effects on both D_LCO and D_LNO, which were lower in Grenoble than in Bordeaux (mean differences 8.5% and 13.2%, respectively). In figure 2 we have used their equations based on the entire population, which therefore results in higher values than if the Grenoble equations had been used. In addition, although they performed at least two acceptable tests for each subject, apparently Aguilaniu et al. [14] used only values from the test with the greater D_LCO, which differs from the ATS/ERS recommendation, according to which the mean of two acceptable measurements should be reported [20]. In the present study we used the mean of two acceptable measurements, and this difference in procedures contribute to the observed discrepancy between the studies. Other possible explanations of the observed differences might be that different equipment was used, as well as differences in the populations studied. Finally, deviations in the simultaneously measured V_A might have a rather large impact on D_LCO and D_LNO. But as Aguilaniu et al. [14] did not present information about V_A, comparison with V_A from the present study and its potential influence on the other presented values cannot be made.

As seen in figure 2, only the predicted values from the present study, the study by Aguilaniu et al. [14] and the combined dataset by Zavorsky et al. [15] take into account the observed accelerated loss of diffusing capacity with age. The reason why Zavorsky et al. [13] and van der Lee et al. [12] have only one slope, when values are fitted for age, might be that the number of old people in their studies have been too low to reliably detect this accelerated change with age. In the case of van der Lee et al. [12], the slopes depicted in figure 2 are markedly less steep than those from the other studies, which is probably also an effect of the relatively young population in that study. Also notable is the rather high value of 5.1 for the D_LNO/D_LCO ratio found by Zavorsky et al. [13] (current consensus is that the ratio is in the range of 4.3–4.9 [6]; values from present study = 4.4, van der Lee et al. [12]= 4.5 and Aguilaniu et al. [14]= 4.75). If this is not a consequence of actual differences between different study populations, it can result from measures of D_LNO being too high or measures of D_LCO being too low or a mixture of the two. We like to think that the first possibility has had the greatest significance, since Zavorsky et al.'s [15] combined values for D_LCO fit almost perfectly with the values from the present study. It should be mentioned that methods, equipment and of course study populations were different in all four studies discussed. In relation to differences in study population, it has been shown that differences in physical activity status have an impact on diffusion parameters [31–34]. Likewise, differences in exposure to air pollution might affect lung function. These aspects have not been analysed in the present study, but they might explain some of the observed variation between studies.

D_LCONO method versus standard D_LCO method

As seen in figure 3, some differences can be observed between values obtained from the D_LCONO method and the standard D_LCO method.

The largest difference is in V_A, with V_A5s being generally lower than V_A10s. In part, this difference might be a result of inadequate mixing of the inert gas with the alveolar gas since short breath-hold times have been shown to lower the measured V_A in some patient groups and in healthy subjects [35–37]. Other important possible reasons for the observed difference are the differences in methodology between the two methods (see table 1). For example, the inert gas used in the calculation of V_A is not the same (He versus CH₄). The two gases might have different distributions in the lung and different solubility in tissue owing to their physical properties, and this might lead to differences in the measured V_A.

K_CO has been shown to increase with decreasing breath-hold time [37]. When looking at D_LCO, this increase in K_CO will tend to counteract the effect of a decreasing V_A on D_LCO. Indeed, classically D_LCO is thought to increase with decreasing breath-hold time, which is shown in studies where breath-hold time is the only factor being changed (that is, same methodology in all other aspects) [37, 38]. In the present study, K_CO5s is generally larger than K_CO10s, as seen in figure 3b. And, as described above, this increase is seen to “compensate” for the decrease in V_A, thereby resulting in D_LCO5s being slightly but significantly larger than D_LCO10s (mean±sd difference=0.85±2.3 mL·min⁻¹·mmHg⁻¹). In summary, as seen in table 1 the two methods differ in a number of ways, and more research is needed in order to determine how these differences in methodology influence V_A, K_CO and D_LCO. What is certain is that D_LCO measured using the two different methods cannot be used interchangeably, that is, specific reference material has to be used for each of the two methodologies.

Sex difference in V_c/V_A, D_mCO/V_A and D_LNO/D_LCO5s

Both V_c/V_A and D_mCO/V_A were to some extend affected by sex, although in opposite directions: V_c/V_A was generally slightly higher in females, while D_mCO/V_A was lower. This suggests that there is a sex difference both in the alveolocapillary membrane and in the pulmonary capillary volume when normalised to V_A. However, this observed sex difference is affected by the method used for correcting for Hb. This should be kept in mind if subsequent studies are to compare similar results with the results presented here.

In contrast, D_LNO/D_LCO5s showed no sex difference. However, it is important to note that the D_LCO5s values used in this calculation were not corrected for Hb. Since Hb is generally lower in females, the resultant values for D_LCO5s should be lower for this reason alone. Therefore, if no sex difference existed between the alveolocapillary membrane and the pulmonary capillary volume when normalised to V_A, then we would expect D_LNO/D_LCO5s to be higher in females than in males, which was not the case in the present study.

D_LNO/D_LCO

It has been pointed out that the D_LNO/D_LCO ratio might be the best way to assess the relationship between D_mCO and V_c. The main argument has been the former lack of consensus regarding the true values of θ_CO, θ_NO and α used in the calculation of D_mCO and V_c, since the D_LNO/D_LCO ratio has the advantage of being independent of these values [6]. Certainly, the ratio can tell us something about the relationship between D_mCO and V_c, and in the case of a low measured D_LCO value it could point in the direction of the parameter (D_mCO or V_c) predominantly accountable for the decrease. However, caution should be exercised when looking at the ratio alone, since an apparently normal value could result from both D_LNO and D_LCO being low, and in addition a low ratio could of course either result from a low value of D_LNO or a high value of D_LCO, while the opposite could apply to a high ratio. Furthermore, the scatter of the normal values for the ratio is rather large (mean±sd 4.4±0.24), and for patients the scatter of values is also found to be large [39, 40]. Obviously, this might result in difficulties differentiating between normal and pathological values. The usage of specific values for D_mCO and V_c could overcome some of these challenges and in addition it could provide a more detailed view of the resistances associated with lung diffusion. But as mentioned, if this is to become reality, consensus has to be made regarding the calculation of D_mCO and V_c. As discussed later, this might be achievable today.

θ_CO, θ_NO and α

In recent years, most scientists have agreed that the most correct values for θ_CO are those presented by Forster [11] in 1987, and thereby not the 1957 values presented by Roughton and Forster [2]. Forster himself argued that these new values were more correct, particularly since they were measured at a physiological pH of 7.4 and not pH 8.0 like the 1957 values [9, 12, 14]. In 2016, Guénard et al. [41] tested several of the available 1/θ_CO versus P_capO2 equations by exposing 10 normal subjects to two different inspiratory oxygen concentrations while measuring D_LNO and D_LCO. Several of the equations managed to keep changes in the D_m/V_c ratio at a minimum during changes in P_capO2, among these the equation proposed by Forster [11]. On the basis of these results, Guénard et al. also proposed a new “best-fit” equation. This equation is used in the work by Zavorsky et al. [15], since they found that there is still insufficient information to decisively choose between the existing published 1/θ_CO versus P_capO2 equations derived in vitro. However, it is important to note that very little difference is seen in values for D_m and V_c when comparing this new equation to the equation by Forster. Therefore, in the present study we have decided to continue with the in vitro Forster equation.

Concerning the true value of α, in line with most other researchers we consider the true value to be 1.97, since this is the theoretical value representing the relationship between the physical solubilities of NO and CO in plasma taking into account their molecular weight [3, 25]. Some researchers have forced α to higher values in order to achieve a better fit of D_mCO and V_c values obtained from the D_LCONO method with values obtained from the oxygen two-step Roughton–Forster D_LCO method. For example, in this way Tamhane et al. [42] found D_LNO/D_mCO (two-step)=2.42. An explanation for this might be that in their calculations they used θ_NO=infinite (thereby assuming D_LNO=D_mNO) together with the 1957 values for θ_CO. If instead they had used θ_NO=finite, D_mNO would not be equal to D_LNO, but would exceed this value by ∼70–80% (according to values from the present study and Hughes and Bates [43]). This would lead to an apparent D_mNO/D_mCO ∼4.1–4.3. However, using 1987 values for θ_CO increases D_mCO approximately two-fold compared to the 1957 values, thereby leading to α values that might be better in concordance with the theoretical value of 1.97 [43]. In any case, since α is defined as the physical diffusivity ratio between NO and CO, the approach by Tamhane et al. [42] cannot be correct.

Much debate has been focused on the correct value of θ_NO. In 1987, Guenard et al. [3] assumed 1/θ_NO to be negligible when they first introduced the single-breath D_LCONO measurement as a possible means of determining D_mCO and V_c. Since then, many researchers have regarded θ_NO as being infinitely great with reference to the very fast reaction rate of NO with free Hb. However, in recent years experiments in vitro as well as in vivo conducted by Borland and colleagues [7, 44] have consolidated the in vitro value of θ_NO=4.5 mL_NO·mL_blood⁻¹·min⁻¹·mmHg⁻¹ first presented by Carlsen and Comroe Jr [45] in 1958. In addition, Borland and colleagues [7, 44] showed that red blood cell lysis in a membrane oxygenator model of CO and NO transfer or substitution of red blood cells with cell-free haem-based oxyglobin in anaesthetised dogs increased D_LNO considerably while D_LCO hardly changed. This has led researchers who previously regarded θ_NO to be infinite to consider it finite with a value of 4.5 mL_NO·mL_blood⁻¹·min⁻¹·mmHg⁻¹ [4, 6, 39].

According to the work of Roughton and Forster from 1957 [2], the red blood cell fraction of the total resistance to CO uptake was estimated to be ∼50%. However, newer evidence including morphometric measurements of D_m and re-calculation of values obtained from the oxygen two-step Roughton–Forster D_LCO method suggests that this fraction is more likely to be ∼75–80% [43, 46]. Some well-known features of D_LCO argue in favour of the view that 1/(θ_CO·V_c) should be the most important rate-limiting factor for CO transfer: 1) anaemia, increase of carboxyhaemoglobin and/or raising of P_capO2 all lower D_LCO; and 2) D_LCO is low in some pulmonary vascular conditions with normal vital capacity [43, 47, 48]. Values from the present study support this more current view of the distribution of resistances for CO uptake, since the average red blood cell fraction of the total resistance in our study was 72.3%. In addition, the same fraction for NO uptake was 39.3%, which parallels the value of 37% presented by Borland et al. [7].

Breath-hold time

Another area that needs to be consistent between studies using the D_LCONO measurement is the breath-hold time. Standard breath-hold time for the D_LCO measurement is 10 s, but this is not suitable for the combined D_LCONO measurement, since NO transfer is ∼4.5 times faster than for CO. This leads to very low concentrations of NO after 10 s, which is therefore undetectable by electrochemical cells. Use of a more sensitive chemiluminescent analyser circumvents this problem, but adds considerably to the expense. In the present study we chose a breath-hold time of 5 s (true apnoea period), which is in concordance with earlier studies [13, 14, 49].

Caution with automated procedures

The data presented have been obtained by using of equipment and largely automated procedures. This has some obvious advantages regarding effectiveness and ease of use. However, since we experienced more than one incident where these automated procedures did not comply with our needs and where manual correction of data therefore was needed, we would like to call attention to the fact that caution has to be taken when using such automated procedures.

Clinical implications

To date, several studies have pointed at the added value of D_LCONO compared to measurement of D_LCO alone when examining patients with different pulmonary disorders. This ranges from pulmonary vascular diseases such as chronic thromboembolic pulmonary hypertension to sarcoidosis and cystic fibrosis [40, 50, 51]. Unfortunately, a lack of concordance concerning the D_LCONO method and computation of D_mCO and V_c complicates the interpretation and in particular, comparison of results. As proposed by Hughes and van der Lee [6], a way to circumvent some of these discrepancies is to look mainly at the ratio of D_LNO/D_LCO, which according to the studies mentioned shows alterations specific to different pulmonary disorders. However, being able to reliably measure D_mCO and V_c and by comparing the results between studies and with the updated reference material presented in this study, we hope that future studies will be able to provide more information on the pathoanatomy and pathophysiology of pulmonary disorders. It seems achievable to use information obtained from the D_LCONO measurement in the everyday clinical work-up of patients.

Conclusion

The present study is one of the largest to date to present reference equations for the D_LCONO measurement. In particular, subjects >70 years of age are very well represented, which is exceedingly important as an increasing number of patients are in this age group. In addition, it is the first large-scale standalone study performed on a single uniform population to present reference equations for D_mCO and V_c derived from the D_LCONO measurement and using current state-of-the-art methodology in the computation of these two measures.

We found age, sex, height and age squared to be independent explanatory variables of the main outcomes. However, the four explanatory variables were not independent predictors of all outcomes. For all outcomes, we found an accelerated loss of capacity with age, which is represented by a negative value of the parameter for the independent variable age squared present in all the reference equations.

We believe that the D_LCONO measurement and its ability to determine D_mCO and V_c has great potential in future research and diagnostics of pulmonary disorders. Yet, in order to reap the full benefits of this technique, in addition to reliable reference equations, consensus concerning methods and computations must be reached. In recent years, much has changed in this field, but finally agreement seems to be within arm's reach. Therefore, we urge future studies to use this newest methodology as it is presented in this article.