Introduction
It has been in 1915 when Marie Krogh developed a method to measure the single-breath uptake of carbon monoxide in the lungs [1]
[2]. Ever since, the measurement of diffusing capacity as it is called in North America or transfer factor as more appropriately referred to in central Europe is a frequently performed measurement and considered as an important tool in medical surveillance examinations of diverse cardiopulmonary diseases [3]
[4]
[5]
[6]
[7]
[8]
[9]. Furthermore, these parameters seem to be of prognostic value in the field of lung resection and transplantation [10]
[11]
[12]
[13]
[14].
The interpretation of diffusing capacity parameters usually relies on the comparison to reference values derived from individuals without present or previous conditions affecting ventilatory function [15]
[16]
[17]. Discrepancies in predicted values among different authors can be found [18]
[19]
[20]
[21]
[22]. These differences might be, above all, explained by different selection criteria of subjects, as well as by methodological and technical differences. A comprehensive list of published reference equations for parameters of diffusing capacity has been published in 1981 by Crapo et al. [18] first. Guidelines for the standardization of measurement of diffusing capacity and a set of reference equations were later published by Cotes et al. in 1993 [23]. Later available studies on parameters of diffusing capacity have been performed on a maximum of 1000 participants in total, with an underrepresentation of elderly people. Therefore, additional results on even larger samples are needed. The ATS (American Thoracic Society) and ERS (European Respiratory Society) guidelines for the measurement of diffusing capacity were updated in 2017 [24]. However, this huge data set based on a retrospective and pooled analysis of 19 centres. Based on this, the aim of our study was to derive a comprehensive, consistent set of prediction equations for single-breath diffusing capacity from a large-scale, population-based cohort in central Europe and to compare the results with those of existing prediction equations.
Methods
Study volunteers comprised participants of the Study of Health in West Pomerania (SHIP), a population-based survey in the northeast of Germany. Study details are given elsewhere [25]
[26]. Data of SHIP-2 and SHIP-TREND were taken for the current analyses. Measurement of diffusing capacity was performed on 3566 individuals aged 20 – 84 years ([Fig. 1]). All participants were investigated in health examination centres established for the purpose of the study and gave written informed consent. The study conformed to the principles of the Declaration of Helsinki as reflected by the approval by the Ethics Committee of the University of Greifswald.
Fig. 1 Age distribution of the study sample.
Socio-demographic and behavioural characteristics as well as information on medical history were collected using computer-assisted personal interviews administered by trained and certified staff. Previous medical history was based on physicians’ diagnoses as reported by the subject. The definition of cardiopulmonary disorders was additionally based on the use of specific medication (ATC code R03). Medication was recorded according to the Anatomical Therapeutic Chemical (ATC) classification [27], and drugs of interest (ATC code R03) were treated as binary variables (0 = no use vs. 1 = use). Smoking status (current, former, never-smokers) and physical activity (no or less 1 h/week, 1 – 2 h/week, ≥ 2 h/week) were assessed by self-report. Weight and height were measured in a standardised manner. Moreover, two-dimensional and M-mode echocardiography were performed using a Vingmed CFM 800A system (GE Medical Systems, Waukesha, Wisconsin, USA).
All clinical tests were performed by experienced, trained and certified staff. Initial certification was awarded to observers after a minimum of 3 months of training. Observers were held to strict quality criteria. To facilitate comparability between SHIP and other population-based studies in Germany, external observers were regularly invited to participate in certification procedures. The data collection phase was monitored by a Data Safety and Monitoring Committee.
Of the 3566 participants with diffusing capacity parameters availability, 1316 subjects were excluded due to the presence of at least one of the following conditions (overlaps existing): self-reported lung disease (n = 174), a FEV1-FVC-ratio < 70 % (n = 254), asthma bronchiale (n = 94), self-reported dyspnea or weakness under physical load, an ejection fraction < 45 % (n = 49), invalid results (n = 5), or history of myocardial infarction (n = 86), heart surgery (n = 45), stroke (n = 52) or heart failure (n = 93). In addition, 473 current smokers and 152 subjects who received medication with potential influence on ventilatory function (ATC code R03) were excluded. Altogether, the final study population for the present analyses consisted of 1786 subjects ([Fig. 2]).
Fig. 2 Selection of the study population.
The examinations were conducted using a variable pressure bodyplethysmograph equipped with a pneumotachograph (VIASYS Healthcare, MasterScreen Body/Diff., JAEGER, Hoechberg, Germany) which met the American Thoracic Society (ATS) criteria [28]. The volume signal was calibrated with a 3.0 litre syringe connected to the pneumotachograph, in accordance with the manufacturerʼs recommendations and at least once daily. Barometric pressure, temperature and relative humidity were registered every morning. Volume calibration referred to ATP-conditions (Ambient Temperature Pressure) but resulting lung volumes were expressed as BTPS-corrected (Body Temperature Pressure Saturated) [28]
[29].
The tests were carried out in accordance to ATS and ERS (European Respiratory Society) recommendations [30] in the following order: (1) determination of static lung volumes, (2) forced spirometry, (3) single-breath CO-diffusion measurement. The procedures were conducted in a sitting position while subjects wearing a noseclip. The procedure was continuously monitored by a physician. Prior to the test the required manoeuvres were demonstrated by the operator, and the individuals were encouraged and supervised throughout the testing. The manoeuvres required for the measurement of diffusing capacity were exercised in a so called training-phase and at least repeated twice.
The variables of interest for the present study were:
-
TLCO: transfer factor of the lung for carbon monoxide;
-
TLCOc: transfer factor of the lung for carbon monoxide, corrected for hemoglobin;
-
TLCO/VA (KCO): transfer coefficient for carbon monoxide (Krogh-Index);
-
TLCOc/VA: transfer coefficient for carbon monoxide, corrected for hemoglobin.
Stratified by sex we reported categorical variables as percentages and continuous variables as median, 25th and 75th percentile. Differences were tested by χ2-test (categorical data) or Wilcoxon test (continuous data). Reference intervals for the diffusion parameters were established stratified by sex by quantile regressions for the 2.5th and 97.5th percentiles with age, height, weight and current smoking as explanatory variables. Equations for the mean of the diffusion parameters were generated by linear regressions with age, height, weight and current smoking as explanatory variables. The values for hemoglobin were available and integrated into the analyses performed for the TLCOc and TLCOc/VA. The resulting equations for the 2.5th percentile and the 97.5th percentile were compared visually with those of the Global Lung Function Initiative in a graphic with age on the X axis and the respective diffusion parameter on the Y axis [31]. The curves show the values of the respective diffusion parameters for the upper and lower limits from our and the GLI formula over age with height being fixed to 177 cm in males and 164 cm in females and weight being fixed to 87 kg in males and 70 kg in females. For comparative analyses only lifelong-non-smokers were included. The GLI limits are defined as 5th and 95th percentiles whereas our limits are 2.5th and 97.5th percentiles. In all analyses a p < 0.05 was considered as statistically significant. All analyses were carried out with Stata 15.1 (Stata Corporation, College Station, TX, USA).
Results
When comparing the characteristics of participants in single-breath diffusion measurement (n = 3566) with those of non-participants (n = 3187), no significant differences in median between the two groups with respect to age (54 in participants vs. 55 years) and BMI (27.5 vs. 27.7 kg/m2) were observed. Participants were less often smokers (21.2 % vs. 28.9 %) and of female sex (49.2 % vs. 55.1 %) than non-participants.
In the next step, the group of participants was analysed with regard to general characteristics ([Table 1]). Men were found to be more often former smokers than women. All parameters of diffusing capacity were significantly higher in males.
Table 1
Descriptive statistics of the study sample.
|
Men
(N = 923)
|
Women
(N = 863)
|
P
|
|
Age (years)[†]
|
54 (43; 65)
|
53 (42; 63)
|
0.60
|
|
Smoking (%)
|
< 0.01
|
|
|
41.17
|
59.68
|
|
|
|
58.83
|
40.32
|
|
|
Weight (kg)
|
87 (79; 96)
|
69 (62; 78)
|
< 0.01
|
|
Height (cm)
|
177 (172; 181)
|
164 (159; 168)
|
< 0.01
|
|
TLCO (mmol/min/kPa)
|
9.41 (8.29; 10.64 )
|
6.87 (6.08; 7.61)
|
< 0.01
|
|
TLCO/VA (KCO) (mmol/min/kPa/L)
|
1.47 (1.33; 1.61)
|
1.43 (1.30; 1.55)
|
< 0.01
|
|
TLCOc (mmol/min/kPa)
|
9.09 (8.04; 10.25)
|
6.63 (5.86; 7.35)
|
< 0.01
|
|
TLCOc/VA(mmol/min/kPa/L)
|
1.42 (1.28; 1.55)
|
1.38 (1.26; 1.51)
|
< 0.01
|
Continuous data are expressed as median (25th and 75th percentiles; nominal data are given as percentages. Differences were tested by χ2-test (categorical data) or Wilcoxon test (continuous data).
† Age at core examination date. For abbreviations please see the method section.
Sub-group analyses among active (n = 473), former (n = 891) and lifelong-non-smokers (n = 895) reveal significant differences. Lower values for all parameters of diffusing capacity can be found in active smokers. Therefore, only non- and former smokers were included in the final analyses. A decline with age can be found for all parameters ([Fig. 3]). The influence of height and weight is obvious. Therefore, the prediction equations were adjusted for age, height, weight as well as former smoking and separately presented for men and women ([Table 2]).
Fig. 3 Age dependency of 2.5 and 97.5 percentiles of TLCO and KCO in men and women, all non- or former smoker.
Table 2
Prediction equations for lung diffusing capacity for CO in men and women.
|
Outcome
|
Parameter
|
Males
|
Females
|
|
TLCO; mmol/min/kPa
|
2.5th perc.
Mean
97.5th perc.
|
– 4.9778 – 0.0542*A + 0.0021*W + 0.0857*H – 0.4372*FS
– 3.3139 – 0.0598*A + 0.0079*W + 0.0870*H – 0.1262*FS
– 8.7631 – 0.0320*A + 0.0106*W + 0.1262*H – 0.6999*FS
|
– 1.7940 – 0.0461*A – 0.0097*W + 0.0620*H – 0.1407*FS
– 3.1956 – 0.0395*A + 0.0034*W + 0.0725*H – 0.0282*FS
– 3.2645 – 0.0347*A – 0.0022*W + 0.0845*H – 0.0271*FS
|
|
TLCOc; mmol/min/kPa
|
2.5th perc.
Mean
97.5th perc.
|
– 5.4157 – 0.0535*A – 0.0048*W + 0.0909*H – 0.6524*FS
– 3.4277 – 0.0541*A + 0.0049*W + 0.0854*H – 0.1134*FS
– 9.3471 – 0.0330*A + 0.0001*W + 0.1308*H – 0.4238*FS
|
– 0.5684 – 0.0460*A – 0.0110*W + 0.0539*H – 0.1064*FS
– 3.1107 – 0.0393*A + 0.0008*W + 0.0717*H – 0.0392*FS
– 5.5505 – 0.0265*A – 0.0035*W + 0.0953*H + 0.1221*FS
|
|
KCO; mmol/min/kPa/L
|
2.5th perc.
Mean
97.5th perc.
|
1.6243 – 0.0076*A + 0.0028*W – 0.0015*H – 0.0694*FS
2.8319 – 0.0091*A + 0.0036*W – 0.0066*H – 0.0286*FS
3.2908 – 0.0086*A + 0.0040*W – 0.0078*H + 0.0077*FS
|
2.5465 – 0.0086*A + 0.0047*W – 0.0079*H – 0.0315*FS
2.4066 – 0.0079*A + 0.0040*W – 0.0051*H – 0.0239*FS
2.2002 – 0.0087*A + 0.0075*W – 0.0032*H + 0.0237*FS
|
|
KCOc; mmol/min/kPa/L
|
2.5th perc.
Mean
97.5th perc.
|
1.2861 – 0.0067*A + 0.0027*W – 0.1177*FS
2.6893 – 0.0082*A + 0.0030*W – 0.0061*H – 0.0263*FS
3.2734 – 0.0089*A + 0.0030*W – 0.0076*H + 0.0477*FS
|
1.8619 – 0.0076*A + 0.0020*W – 0.0029*H – 0.0582*FS
2.3158 – 0.0079*A + 0.0034*W – 0.0046*H – 0.0252*FS
2.4706 – 0.0092*A + 0.0064*W – 0.0043*H – 0.0252*FS
|
A = Age in years, W = Weight in kg, H = Height in cm, FS (Former Smoking) = 1, Lifelong-non-smoking = 0
The visual comparison of the present prediction equations with previously reported series from the Global Lung Function Initiative [31] reveals good comparability for the upper and lower limits of normal ([Fig. 4]).
Fig. 4 Age dependency of upper and lower limits for TLCO and KCO for a (left) man (177 cm, 87 kg) and a (right) woman (164 cm, 70 kg), all lifelong-non-smoker. Comparison between SHIP-data and GLI-data.
Discussion
To the best of our knowledge the present study is one of the largest prediction equations deriving study for parameters of diffusing capacity in a representative sample of healthy adults using the platform of the population-based Study of Health in Pomerania. Individuals covered wide ranges of age, height and weight. By excluding subjects with cardiopulmonary disorders and current smoking an adequate population for reference purposes was created. The mean age of the present study group was 53 years for both gender.
Patients seen in a clinical lung function laboratory are usually in a middle age or older age group. This aspect makes the present prediction equations clinically relevant as parenchymal and pulmonary vascular lung diseases, that make the measurement of TLCO necessary, predominantly occur in middle age and older people.
In comparison to previously reported prediction data, the present data show reasonable comparability [31]. However, previous studies on normative equations of diffusing capacity parameters [18]
[19]
[20]
[21]
[22] either included fewer subjects [32] or were focused on younger individuals [16]
[33]. The application of prediction equations to patients with an age range other than the one included in a specific study, may lead to considerable error. The age range of the present study is 20 – 84 years which makes the established prediction equations fairly good applicable.
The graphical comparison with the GLI prediction values ([Fig. 4]) reveals good comparability for the lower and upper limits of normal range. Whereas the GLI data indicate the 5th and 95th percentile range our data deliver the 2.5th and 97.5th percentile range. This makes our data even more distinct and underlines the excellent characterisation of the healthy study sample.
There is large controversy in the literature regarding the inclusion of former-smokers. Half of the individuals included in the present analyses were classified as former smokers (n = 891). The influence of smoking status on the parameters of diffusing capacity was statistically tested. Being a former-smoker was a significant predictor for the parameters of diffusing capacity, therefore the equations were adjusted for this circumstance and former-smokers included in the current analyses. This aspect should improve the generalisability of the current prediction equations as a high proportion of patients presenting in a lung function laboratory are former smokers. The ability to adjust for this factor improves the likelihood to detect pulmonary diseases. A similar approach has been performed with the establishment of normative values for cardiopulmonary exercise testing on the SHIP sample [34]
[35].
Weight is a significant predictor on parameters of diffusing capacity in the present analyses, as firstly shown by Roca et al. [36]. However, the weight term may falsely elevate the predicted parameters of diffusing capacity in overweight and obese individuals. The mean weight of the present population was 79 kg with a relatively wide range from 45 kg to 159 kg. This should make the bias negligible.
Similar to prediction equations for spirometry [37], gender, age and height were independent predictors for TLCO. The present prediction equations are therefore sex-specific and describe a multiplicative relationship with age, weight and height.
The equations obtained differ from those previously reported not only in their mathematical form but also in the type of underlying data. The former ERS reference equations for example were linear and obtained by summarizing published regression equations from older surveys published between 1950 and 1980 using different equipment [38]. Quality control and standardization according to ATS/ERS guidelines may partly explain differences with other studies. Older publications report TLCO values based on outdated equipment, using different gas concentrations or algorithm. However, our study is up to date, using modern equipment and methods that are in accordance with the ATS/ERS guidelines. Our results are in good agreement with current studies, especially with the global reference values for TLCO derived from the Global Lung Function Initiative in 2017 [22]
[31]
[39]. The GLI (30) collected data from all normal value papers published since 2000 and analysed n = 9710 subjects. Therefore it is much greater than our data pool. However, a pooled analysis cannot apply such strict inclusion criteria as our study, e. g. echocardiography or structured interview. Therefore, the SHIP study is far more homogenous and might have the more precise values for the middle European population.
The present data should not be applied to non-Caucasians as a possible influence of ethnicity on parameters of lung function is discussed in the literature [31]
[40]. Moreover, no chest radiographical examinations were available. Therefore we cannot completely exclude that some patients might have had unknown and unreported asymptomatic lung disease. It is, however, very likely that this would have affected only a very small proportion of data. Furthermore, due to voluntariness a selection bias cannot completely be ruled out. However, we did not find significant differences for age and BMI between participants and non-participants in lung function measurement.
The measurement of diffusing capacity is dependent on the subjectʼs performance and examiners instructions. Due to an initial training phase and at least two measurements in accordance with the ATS/ERS guidelines [24] this bias should be insignificant. There may have been under-reporting of previous smoking or even other confounders such as passive smoking or occupational exposure. As reported in previous epidemiological analyses on lung function (EPIC-Norfolk Population) there seems to be a well described influence of the social status on the parameters of lung function [41]
[42]. However, this was not part of the present analyses.
The presented reference values for the measurement of diffusing capacity may contribute to the updating of old existing reference values by respiratory societies especially in populations with Caucasian characteristics.
