Keywords

1 Introduction

Basic prerequisite of any organism is the oxygen supply to all its cells. The central organ of gas transportation is the lung and the blood is a carrier. Oxygenation is achieved by the ventilatory gas exchange and diffusion of gases through the alveolar and capillary walls. The measurement and determination of gas exchange in the lung is therefore a key requirement in the diagnosis of vital functions, especially in obstructive and restrictive lung diseases.

Commonly used is the measurement of the diffusing capacity for CO (DLCO); CO diffusion having the characteristics similar to those of O2 diffusion. An alternative is a measurement of the results of gas exchange, namely the content of O2 and CO2 in arterial or arterialized capillary blood. The diffusion capacity of the lung for gas represents an integral of respiratory function since ventilation, diffusion, and perfusion are included in the measurement.

In general, restrictive lung disease, like fibrosis, results in decreases of oxygen diffusion and arterial oxygen partial pressure (PaO2), especially during exercise, compared with obstructive airway diseases which are often associated with a mismatch of ventilation and perfusion. Although both diseases show the key symptom of dyspnea, the cause of dyspnea may be diverse.

Both hyperventilation and hypoventilation affect the pulmonary uptake of O2 and consequently the PaO2: hyperventilation leads to an increase of PaO2, so that a malfunction of gas exchange may be underestimated. Therefore, determination of diffusion characteristics through blood gas analysis should take into account the level of ventilation. This integration of ventilation in the assessment of blood gases is enabled by determining the alveolar-arterial oxygen difference (PA-aO2). It requires not only the measurement of blood gas content but also of the breathing gases O2 and CO2 by means of a complex technique. This methodology is always part of a cardiopulmonary exercise testing (CPX).

Another way to calculate the influence of hyperventilation on the PaO2 is to assess a decrease in the arterial partial pressure of carbon dioxide (PaCO2). The PaCO2 is here used to quantify the influence of ventilation; the PaO2 can thus be ‘corrected’ according to the formula: PaO2corr = PaO2 − 1.66 × (40-PaCO2) (Diekmann and Smidt 1984). The mathematical correction of PaO2 is much simpler to perform than the determination of the PA-aO2. The question arises whether the results of these two measurements are equivalent. Therefore, in the present study we investigated the validity of PaO2, PaO2corr, and PA-aO2 – all in comparison to DLCO in various lung diseases.

2 Methods

2.1 Participants

The participants were recruited in our occupational outpatient clinic over a period of 5 years and all of them gave signed written consent to use their samples and data. The study was approved by the Internal Medical Review Board. They presented themselves for the diagnosis of work-related diseases and occupational medical examinations. A total of 250 subjects (mean age 52.3 ± 12.5 year; F/M 40/210) were eligible and consecutively included after they had performed CPX with blood gas analysis, DLCO, or both as part of their routine investigation. In addition, medical history, physical examination, spirometry, and body plethysmography were taken in all subjects. We excluded 10 cases due to single missing values. From the remaining 240 individuals, there were: 13 with restrictive lung disorder, defined as VC < lower limit of normal (LLN) (with normal FEV1/VC, DLCO normal or reduced); 19 with normal VC but DLCO < LLN; 86 with mild or moderate bronchial obstruction (FEV1/VC < LLN, VC > LLN), 34 subjects with a mixed obstructive/restrictive lung disorder (FEV1/VC < LLN, VC < LLN), and 88 healthy controls without past or present pulmonary disorders and with normal lung function results. The examinations were performed as part of routine social security screening.

2.2 Lung Function Tests

Lung function and CPX tests were carried out according to the quality criteria of the European Respiratory Society (ERS) and the American Thoracic Society (ATS) (Meyer et al. 2013; Macintyre et al. 2005; Miller et al. 2005; American Thoracic Society 2003). Blood gas analysis at rest was based on the target values of Woitowitz et al. (1969) and the DLCO on those of Cotes et al. (1993). The PaO2 during exercise was deemed pathological if it fell below the predicted value by ≥5 mmHg (Meyer et al. 2013). The PA-aO2 was calculated at rest and under load from the measured values of CPX using the formula: PAO2 = FiO2•713-(PACO2/respiratory exchange rate (RER)) (Riley and Cournand 1949). A pathological increase was assumed at a value of >20 mmHg at rest and >35 mmHg during exercise (Meyer et al. 2013; American Thoracic Society 2003).

2.3 Statistical Analysis

Pearson’s correlation coefficient (r) was analyzed for DLCO with PaO2, DLCO with PaO2 after correction of ventilation and DLCO with PA-aO2, all measured at rest and under load. Correlations of blood gas-dependent parameters (PaO2, PaO2corr, and PA-aO2) were determined with each other; all calculations were carried out separately for each group of lung diseases.

Cohen’s kappa coefficient (Grouven et al. 2007; Thompson and Walter 1988) was calculated to assess the conformity of the values measured with different methods; where ‘1’ indicates a full match, ‘0’ indicates a purely random coincidence, and negative values represent an even lower than a random match. Crosstabs were made to compare the quality of different measurement methods (healthy/pathological assessments). In 39 male subjects, DLCO values were evaluated as based on the level of current hemoglobin concentration corrected (Mottram et al. 1999) and compared with the DLCO of the total cohort. All correlations were calculated according to Pearson (1909), as all variables were interval scaled and normally distributed. Statistical analysis was performed with a commercial SPSS package ver. 19 and 20.

3 Results

The DLCO value (% predicted value) showed in the total cohort only a low correlation of 0.25 (p < 0.001) to PaO2 at rest and a moderate correlation of 0.57 (p < 0.001) to PaO2 during exercise (Fig. 1a, b, Table 1). The measurement of PA-aO2 under load, which takes into account ventilation, showed only a moderate correlation with DLCO of −0.47 (p < 0.001) in the total cohort. This correlation remained at a similar level of 0.44 for the ‘corrected’ PaO2 that takes into account PaCO2 (Table 1). Higher correlations were found in the first two groups of restrictive lung disease (corresponding with reduced VC or normal VC, but reduced DLCO). Poor correlations in the group with normal lung function values (‘healthy lung’) can be explained by the closely adjacent individual values (see dense point clouds of this group in Fig. 1a, b).

Fig. 1
figure 1

Correlation between DLCO (%pred.) and PaO2 at rest (a) and under exercise load (b)

Full size image
Table 1 Correlation coefficients (r) of DLCO with PaO2 at rest and PaO2 under load, and with PA-aO2 and PaO2corr under load
Full size table

The PaO2 and PA-aO2 highly correlated with each other at rest and also under load (r = −0.83, r = −0.83, respectively), which was particularly evident comparing the PaO2corr and PA-aO2 with respect to the ventilation values (r = −0.89, r = −0.92, respectively) (Fig. 2a, b, Table 2). Such high correlations were confirmed by differentiated calculations using crosstabs and kappa values. Kappa values and crosstabs showed a strong concordance, especially under load (κ = 0.69), particularly within the group of restrictive lung disease (r = −0.95, κ = 0.68). Therefore, PA-aO2 offers no diagnostic advantage over the corrected PaO2 (Table 3), wherein this consideration is essentially dependent on the underlying limits of normal.

Fig. 2
figure 2

Correlation of PaO2corr and PA-aO2 at rest (a) and under exercise load (b)

Full size image
Table 2 Correlation coefficients (r) of PA-aO2 with PaO2corr at rest and under load
Full size table
Table 3 Kappa values (κ) for comparison of PaO2corr with PA-aO2 under load: A – in the total cohort and B – in the restrictive lung disease group
Full size table

When not using the specified correction for load of minus 5 mmHg for the lower limit of PaO2 (Meyer et al. 2013), a full match (κ = 1.0) of the crosstabs for the PA-aO2 compared with PaO2corr was present in the group of restrictive lung disease (data not shown). The crosstabs and kappa values for the blood-gas dependent parameters at rest (PaO2corr and PA-aO2) with the DLCO showed weak correlations (DLCO and PaO2corr: r = 0.17, κ = 0.10; DLCO and PA-aO2: r = −0.19, κ = 0.06) in the total cohort.

Blood gas levels in combination with their corresponding parameters from the exercise test showed a moderate correlation to DLCO and – according to the kappa value – low dependence in the clinical assessment (DLCO and PaO2corr under load: r = 0.44, κ = 0.22; DLCO and PA-aO2 under load: r = −0.47, κ = 0.23, see Table 4). Similarities were mainly in the group of persons with restrictive lung disease; even there, significant correlations were present when the blood gas-dependent values were obtained under load (DLCO–PaO2corr load: r = 0.74, κ = 0.24; DLCO–PA-aO2 under load: r = −0.84, κ = 0.41). In the other groups (obstructive airways disease; mixed restrictive/obstructive disorders; unobtrusive lung function), the similarities were rather weak (data not shown).

Table 4 Kappa values (κ) for comparison of DLCO at rest with: A – PaO2corr under load and B – Pa-aO2 under load
Full size table

As a supplement, the difference between the DLCO value corrected to the current level of hemoglobin and the otherwise underlying DLCO value was calculated in 39 male patients (with assumption of a hemoglobin level of 14.6 g/dL). There was a small deviation in the mean DLCO of 3.0 ± 2.2 %. This outpatient study, with the exclusion of severely ill people, showed the influence of the hemoglobin level to be of little relevance.

4 Discussion

Blood gas analysis and DLCO are the most important diagnostic steps in the assessment of pulmonary gas exchange in routine diagnostics. However, both methods are not always available and their results are affected differently by the respective pulmonary disease and hypo- or hyperventilation. The determination of the alveolar-arterial oxygen difference (PA-aO2) allows the inclusion of ventilation in the assessment of blood gases. This, in turn, requires not only the blood gas analysis, but also the determination of the exchange of respiratory gases O2 and CO2 by means of a complex measuring equipment. Such equipment is part of a cardiopulmonary exercise testing and the method is thus used frequently in the context of cardiopulmonary exercise tests. The influence that ventilation exerts on the PaO2 can be assessed by calculating the exhaled CO2 and thus the ‘correction’ of PaO2 can be made. The PaCO2 is used to quantify the influence of ventilation. In practice, the corrected PaO2 is attained with the results of blood gas analysis in combination with a simple ‘correction formula’ (Diekmann and Smidt 1984). Furthermore, we wanted to verify whether the blood gas values allow for the identification of various lung gas exchange disorders, such as – based on an entirely different principle of measurement – determination of diffusion of CO in the lung (DLCO). It should be noted that the DLCO is to be determined only at rest.

The DLCO is a result of two measurements during a single-breath method (Hughes and Pride 2001): the diffusion gradient at the alveolar membrane and the ventilated alveolar volume. Both result from the measurements of volume, gas concentrations, and calculations. The assumption that DLCO correction using the alveolar volume (DLCO/VA) leads to a more accurate determination of lung diffusion capacity cannot be confirmed by recent publications, since the change of the quotient is not constant with the change of alveolar volume (Hughes and Pride 2012). Therefore, just DLCO and not the DLCO/VA ratio was considered in the context of the present work. The DLCO is considered the gold standard to verify lung diffusion disorders, regardless of their genesis. A closer look at our results reveals that this may apply only for restrictive lung diseases, at least in comparison with blood gas analysis, even after correction of ventilation. A decrease in PaO2 found in blood gas analysis also points to this disease, but the severity of gas exchange impairment could be underestimated if hyperventilation is not observed. Therefore, the arithmetical correction of ventilation can also be useful here.

In obstructive airway disease, in 45 % of cases, the CO diffusion disorder cannot be confirmed by blood gas analysis and PA-aO2 during exercise. In our opinion, the DLCO also appears negatively affected by inhomogeneity of ventilation and perfusion resulting in gas exchange disorders even at rest. This inhomogeneity is known, in particular, for obstructive lung diseases. Only exercise tests seem to provide a better differentiation of a fixed diffusion disorder. Schwarz et al. (1999) also concluded that the PA-aO2 determined by CPX is a more sensitive parameter, compared with PaO2, in the evaluation of gas exchange disorders. The authors also found only a weak correlation of DLCO to PA-aO2. Both parameters would have a better match if one would measure not only PaO2, but also DLCO under load. This would counterbalance the ventilation-perfusion inequality. This inhomogeneity in obstructive lung disease appears of less importance in restrictive lung diseases. DLCO measurements in the loading condition would give a truer assessment of the factual gas diffusion – however, this is not yet available. Furthermore, inhomogeneity of lung perfusion and ventilation under load would be reduced and the entire system of gas exchange would be tested at load limit.

At present, DLCO and blood gas content measurements do not provide comparable values enabling their clinical evaluation. To detect malfunctions in the system, the blood gas content under load, determined with a correction of PaO2 in rapport with the level of ventilation, has the best explanatory power for clinical assessment. The PA-aO2 has a similar power, but there are no reliable set point-values, which complicates the clinical evaluation. Thus, PA-aO2 shows no significant advantage over the PaO2. Nevertheless, one should - as also others report (Schwarz et al. 1999) – use the parameters associated with history, clinical, laboratory values, and imaging techniques. Exercise testing enhances the evaluation of severity, prognosis, and treatment monitoring (Meyer et al. 2013).

The collected absolute values – in particular, the blood gas parameters derived there from PA-aO2 and PaO2corr – are well comparable and highly correlated, so that the explanatory power of the PA-aO2 against the PaO2corr seems diagnostically equivalent. However, given the need to use the CPX system to determine ventilation and PA-aO2 or the above-mentioned calculations to obtain PaO2corr, using the PaCO2, there is a small advantage compared with the determination of PaO2 alone. The main challenge consists of setting a demarcation line between normal and pathological values. This is by far only vaguely defined for PA-aO2 with a limit of 35 mmHg across all age groups and all load levels. The assessment of blood gases also shows discrepancies. The generally assumed limit of 5 mmHg below the normal value is questioned as it may be age- and exercise-dependent.

For the present study, it is essential to note that X-ray images of the subjects investigated were not always present. It was assumed that relevant changes would be reflected in pathological lung function values. It should also be noted that the maximum workload was not defined and was not included in the analysis. The subjects achieved their individual maximum wattage depending on gender, height, weight, age, fitness level, and an existing lung disease. The increase in wattage per minute was selected depending on the expected overall performance and took place after 8–12 min (Preisser and Ochmann 2011). Accordingly, these factors were highly variable. The aim of the study was to capture the effect of blood circulation and pulmonary ventilation as prominently as possible; therefore, the endpoint of maximum workload was selected.

5 Conclusions

In general, only a few conclusions about the existence of a gas exchange disorder can be drawn from the blood gas analysis at rest. The DLCO also is determined only at rest, but shows a slightly better match with the results of the blood gas content during exercise than that at rest. The exercise testing is thus desirable.

Exercise testing is becoming increasingly important in the evaluation of disease severity, prognosis, and therapy monitoring. The inhomogeneity of perfusion and ventilation, influencing the PaO2, can be revealed through the exercise test, not only for obstructive airway diseases but also in healthy subjects (Meyer et al. 2013). The exercise test should include the determination of blood gases, and in the case of cardiopulmonary exercise tests, also the alveolar-arterial oxygen difference. The PA-aO2 takes into account the level of ventilation, thus it probably has a better correlation with DLCO compared with the blood gas analysis.

In restrictive lung disease, all three parameters are comparably suitable to detect the gas exchange disorder. In obstructive airway disease, DLCO seems affected by other pathophysiological aspects; thus there is only a moderate correlation with the blood gas-based parameters. The blood gas analysis at rest can lead to false-negative results, especially in case of restrictive lung disease. The DLCO indicates more likely false-positive results, especially for obstructive airway diseases.

Changes in PA-aO2 and PaO2 during exercise are highly comparable in patients with restrictive lung disease. An exercise test with the determination of blood gases seems to be diagnostically adequate in these cases, at least for the evaluation of gas exchange disorders.

The ‘correction’ formula of Diekmann and Smidt (1984) is also applicable for the blood gas content in the exercise load condition, but gives no advantage over the PaO2 in the assessment of different lung diseases. However, clinical classification as ‘normal’ or ‘pathological’ shows only a moderate difference in PA-aO2 and in ventilation corrected PaO2. The most significant problem in the clinical application of the blood gas parameters at exercise and of PA-aO2 stems from the lack of clearly defined normal-to- pathological range, in particular, for the PA-aO2, where validated reference values are needed.