Journal of the New Zealand Medical Association, 27-October-2006, Vol 119 No 1244
Complete reference ranges for pulmonary function tests from a single New Zealand population
Suzanne Marsh, Sarah Aldington, Mathew Williams, Mark Weatherall, Philippa Shirtcliffe, Amanda McNaughton, Alison Pritchard, Richard Beasley
Reference ranges for pulmonary function tests are a prerequisite for the accurate interpretation of results.1 They are also important in a wide range of respiratory disorders in which guidelines concerning diagnosis, assessment of severity and management rely on lung function measurements expressed in relation to reference values.2,3 The combined American Thoracic Society (ATS) and European Respiratory Society (ERS) guidelines encourage regular review of reference equations suggesting 10 years as an appropriate equation lifespan.4
Reference equations should be derived from study populations closely anthropomorphically and ethnically matched to the populations in whom they are used.4 This implies it would be beneficial to use reference equations derived from the same study population for all parameters. However, few authors have supplied such data,5–8 and although numerous equations have been produced for spirometry, fewer are available for static lung volumes and gas transfer.4 The most recently derived reference equations from a New Zealand population were produced over 25 years ago.9
The widely used equations of the European Community for Steel and Coal (ECSC), first published in 198310 and adopted by the ERS in 1993,11,12 provide a single source of reference ranges for many commonly measured respiratory parameters. These equations have frequently been used in studies of respiratory function13,14 including the European Community Respiratory Health Survey (ECRHS) in which New Zealand played a significant part.15 However, although these equations are presented as a single source, they are composite equations derived from many separate studies conducted in different populations.10
Although validated at the time of their original publication,11 more recent data suggests that at least some of these equations, particularly for spirometry, are now out of date in European populations.16,17 In addition, ECSC reference ranges are not available for newer parameters such as the forced expiratory volume in 6 seconds (FEV6), which may become increasingly used as a substitute for FVC in the detection of airflow obstruction.18,19
To address these issues, we obtained a complete set of pulmonary function tests from a single population of non-smoking New Zealand adults of European origin, free from respiratory disease and symptoms. Reference equations were derived, using linear regression techniques, for all commonly measured respiratory function parameters. Pulmonary function measurements were compared to those predicted by both the ECSC/ERS equations10–12 and more recently published equations for static lung volumes.8
Study participants—Study participants, forming part of the Wellington Respiratory Survey (WRS), were recruited using a postal questionnaire sent to 3500 individuals, aged 25 to 75 years, randomly selected from the electoral register. Subjects returning completed surveys were invited to undertake a more detailed, interviewer-administered, questionnaire followed by pulmonary function tests.
To increase the number of normal participants, healthy subjects were also recruited from a concurrent study investigating the pulmonary effects of marijuana smoking. This represented a convenience sample of adults aged between 18 and 70 years, recruited through newspaper and radio advertisements and informal contacts.
Normal subjects from both studies were defined using ATS guidelines,20 and for this study, were required to self-identify as “New Zealand European”, and be never smokers with no diagnosis of respiratory disease, no recent respiratory symptoms, and no use of inhaled medication (Table 1).
Both studies were approved by Wellington Ethics Committee, and written informed consent was obtained from each subject.
Pulmonary function testing—Pulmonary function tests undertaken as part of the WRS have been described in detail.21 In brief, tests were carried out by one of three trained operators (SA, SM, MVW) at one site using two Jaegar Master Screen Body volume constant plethysmography units with pneumotachograph and diffusion unit (Masterlab 4.5 and 4.6 Erich-Jaegar, Wurzberg, Germany). Prior to testing, equipment was calibrated daily in accordance with ERS and ATS guidelines.22–24
Subjects were not tested within 3 weeks of an upper or lower respiratory tract infection (cough, sputum production, sore throat, or nasal congestion). Subjects were weighed (wearing indoor clothing without shoes) to the nearest kg and height was measured, without shoes, using a vertical backboard, to the nearest 0.5 cm. Subjects greater than 125 kg were excluded due to simultaneous involvement in a study involving CT scanning in which a weight limit was determined by the limits of scanner equipment.
Airway resistance, static lung volumes, and slow spirometric measurements were made prior to forced manoeuvres and gas transfer measurements followed these. A nose clip was worn for all tests.
Airway resistance was measured during relaxed breathing at a rate of approximately 0.5 Hz. Following attainment of a stable baseline, thoracic gas volume representing functional residual capacity (FRC) was calculated using plethysmography. The subject was then instructed to breath out comfortably, maximally inspire, and expire slowly to measure slow vital capacity (SVC).
Expiratory reserve volume (ERV) was measured from FRC to the point of maximum expiration. Inspiratory capacity (IC) was calculated from SVC – ERV and residual volume (RV) was calculated from FRC–ERV. The total lung capacity (TLC) was calculated as SVC (measured) + RV.
FRC and airway resistance measurements were used to calculate specific airway resistance (sRaw) and conductance (sGaw).
Spirometry measurements were carried out in accordance with ATS criteria.22 The largest FEV1 and FVC from a minimum of three acceptable manoeuvres were used with all other flow volume parameters taken from the manoeuvre with the largest combination of FEV1 and FVC.
FEV6 and FEV1/FEV6 measurements were obtained by analysis of stored data.
Gas transfer was measured in accordance with ATS criteria,24 and a minimum of three measurements of diffusion capacity (DLCO) were made with 5-minute intervals. The average DLCO of tests meeting recommended reproducibility criteria were reported.24
Alveolar volume (VA) was simultaneously measured, using helium as a tracer gas, and used to calculate carbon monoxide diffusing capacity per unit of alveolar volume (DLCO/VA).
A blood sample was taken to adjust results for haemoglobin concentration using the formula recommended by the ATS.24 Results for pulmonary function tests except those measuring gas transfer were corrected to BTPS units.
Table 1. Selection criteria for the normal population
†CB: Chronic bronchitis, E: Emphysema, A: Asthma, COPD: Chronic Obstructive Pulmonary Disease, B: Bronchiectasis; *Production of sputum on most days for as much as three months each year; ¶Cough on most days for as much as three months each year; # Did not self-identify as New Zealand European.
Statistical analysis—Reference equations were developed by linear regression. Sex was included as an explanatory variable in each model and for each lung function a further group of candidate explanatory variables was explored; height, height2, age, age2, weight, and body mass index (BMI)—plus interaction terms between sex and age and sex and height. The best regression equation for each parameter was developed with regard to the amount of variability explained, the number of explanatory variables, and consistency between parameters measuring similar physiological aspects of lung function.
R-squared and Mallow’s Cp were used to examine different predictor equations. Mallow’s Cp adjusts R-squared by a penalty term that takes account of the number of predictors in an equation.25 Normality assumptions were tested by examination of the distribution of residuals from the regression equations, by histograms, QQ-plots, and formal tests of normality of the residuals.25
Agreement between the newly developed equations and the ECSC/ERS [known henceforth as European] equations10–12 was by plots of the difference between the actual value of a particular lung function minus its predicted value using each equation (difference plots).
If height or age were important predictors of bias (actual value minus predicted value) then the estimates of it were calculated at the mean age or height. Bias between measured and predicted values was also examined for the static lung volume equations produced by Roca et al.8
SAS (version 8.2) software was used for all statistical analyses.
Initial recruitment from the WRS study resulted in 2319 responses from 3500 screening questionnaires. Of these subjects, 758 completed the detailed questionnaire and pulmonary function tests. The secondary recruitment resulted in 267 subjects completing the detailed questionnaire and pulmonary function testing thus giving a total population group of 1025 subjects.
Exclusion of subjects with previous smoking; diagnosis of a respiratory disorder; symptoms suggestive of COPD, asthma, or bronchiectasis; recent inhaler use; and those who did not self-identify as New Zealand European resulted in 212 subjects in the reference population (Table 1).
The final normal cohort consisted of 180 subjects from the WRS study and 32 from the secondary recruitment. Subject characteristics are shown in Table 2.
Table 2. Subject characteristics
Table 3 shows the lung function reference equations derived from the WRS data. Different intercept terms were created for males and females except for the equations for FEV1/FVC, FEV1/FEV6, FRC, and DLCO/VA in which no significant sex effect was noted. The values for sRaw and sGaw are presented as single expected values for males and females since sex was the only important predictor in this model.
The R-squared (R2) values represent the percentage of explained variance in each measurement accounted for by the equation.
Table 3. Wellington Respiratory Survey (WRS) prediction equations for lung function
Prediction equations take the form:
Abbreviations: FEVn: Forced expiratory volume in n seconds (L); FVC: Forced vital capacity (L); PEF: Peak expiratory flow (L min-1); MMEF: Maximal mid expiratory flow (L min-1); FEFn: Forced expiratory flow rate when n% of FVC expired (L min-1). sGaw: Airway conductance (s-1·kPa1); sRaw: Airway resistance (kPa·s); SVC: Slow vital capacity (L); IC: Inspiratory capacity (L); FRC: Functional residual capacity (L); TLC: Total lung capacity (L); RV: Residual volume (L); ERV: Expiratory reserve volume (L); DLCO: Gas transfer (mmol/min/kPa); VA: Alveolar volume (L).
Most lung function parameters met normality assumptions and did not require transformation. A natural logarithmic transformation appeared to give a better statistical model for PEF (which did not meet normality assumptions) suggesting that the relationship between age, sex, and height was multiplicative for this variable.
Results for maximal mid expiratory flow (MMEF or FEF25-75), sRaw, and sGaw did not meet normality assumptions as well as other lung function measurements but were not improved with logarithmic or other transformations.
The majority of the equations use sex, height, and age as explanatory variables. Residual plots did not suggest that the addition of squared or higher power terms to model curvature improved the fit of any of the equations.
In univariate analyses of the individual variables, weight accounted for a moderate amount of explained variance (data not shown) but was probably acting as a substitute for body size, since, in the multivariate analysis, addition of weight added little to the predictive value of the equations. The only exception to this was for FEV1/FEV6 where weight had a weak effect in the multivariate analysis. Adding BMI was of marginal statistical significance and did not substantially improve the model fit for any equation. The addition of sex by age or height interaction terms was not helpful in developing better prediction equations.
A standard method of calculating normal limits for measured parameters in individual subjects is to add or subtract 1.645 × the root mean square error (RMSE) of the equation from the predicted value17,20,26 and these values are given in the final columns of Table 3.
Comparisons of the WRS equations with the European equations10–12 for FEV1, TLC, RV, and DLCO are shown as difference plots in Figure 1. Table 4 shows a summary of mean bias (measured minus predicted values) by sex.
Predicted values for spirometry using the European equations10,11 were significantly lower than measured values particularly for FEV1 (mean bias, male 0.48 L; female 0.36 L) and FVC (mean bias male and female 0.77 L). In many of the spirometry variables, the difference was greater than 12% of the mean lung measurements of the cohort (Table 4). For FEV1, bias was weakly dependent on age decreasing by 0.05 L per decade.
For lung volumes large differences were seen between measured and predicted values using the European equations10,11 (Figure 1, Table 4). The greatest consistent difference, of around 15% of the mean predicted measurement, was seen in TLC with a mean bias (measured minus predicted value) of 1.14 L for males, 0.89 L for females. Bias was height dependent, being greater in taller subjects (increase in bias (L) per 10 cm increase in height: SVC 0.084, TLC 0.22, FRC 0.23, RV 0.12). The bias for RV was also weakly dependent on age (increased bias of 0.05 L per decade). Measured lung volumes were also compared to those predicted by the equations of Roca et al8 derived using measurements obtained with a plethysmographic technique (Table 5). The bias in these predicted values was significantly less than that seen with the European equations.10,11
Table 4. Comparison between measured values of WRS study and predicted values using the equations of the ECSC/ERS
The columns show the estimated bias (measured minus predicted value) for each parameter when the equations of the ECSC are compared to measured values obtained in the WRS study. Results are shown as mean difference in measured units and as a percentage of the mean measurement for all male and female subjects in the cohort. A positive figure means that the measured value was greater than the predicted value (under-prediction) and a negative value that the measured value was less than the predicted value (over-prediction).
Abbreviations: *The 1983 10 and 1993 11 publications of the formulae for DLCO/VA (also known as KCO) are different giving different values for the mean bias depending on which equation is used.
Table 5: Comparison between measured values of WRS study and predicted values using the equations of Roca et al8
See Tables 3 and 4 for abbreviations and explanation of terminology
Measured results for gas transfer variables show close agreement with those predicted by European equations,10,12 with the only exception being the results for DLCO/VA using the 1983 ECSC formulae.10
The difference plots (Figure 1) show that the variability and spread of the differences between actual and predicted values are similar for both the European equations10–12 and those of the WRS. This suggests that the precision of the equations, reflecting the variability of the sample with respect to predicted values, was similar whichever equation was used.
This study provides reference equations for all commonly used measurements of static and dynamic lung function and new reference values for specific resistance and conductance from a single adult New Zealand population of European origin. In addition, we have produced reference equations for FEV6 and FEV1/FEV6. To the best of our knowledge, this is the first time that reference equations have been derived for this complete set of respiratory function measurements on the same well characterised Caucasian reference population.
Methodological issues—A number of methodological issues relevant to the reporting of these results were identified. Whilst this sample size of 212 subjects was adequate for validation of pre-existing equations,4 it is smaller than that used to derive reference equations in several recently published studies.8,26,27 However other published reference equations dealing with a wide range of testing modalities were obtained using a similar sample size.5
Subjects who did not self-identify as “New Zealand European” were excluded since a low response rate resulted in this group representing less than 7% of those completing testing despite approximately 20% of the local population being non-European in origin.28 The priority will now be to develop comparable reference equations for Maori and Pacific Island people.
The age range of the participants in this study also has implications for the clinical use of the reference equations, being restricted to adult subjects aged 25–75 years. Our equations incorporated sex and standard variables such as age and height. We found that use of polynomial expressions did not improve the fit of our equations.
Although other studies have produced equations incorporating such terms,17,29,30 we have taken a pragmatic approach under the simple assumptions of linear relationships, when they exist, in an attempt not to ‘over fit’ the data. This should make our equations more likely to be appropriate in other populations similar to our own.
Sex is a major determinant of lung function,20 although it is not typically included as an explanatory variable in reference equations. We decided a priori to include this variable since inclusion of a factor accounting for such significant variation in pulmonary function should improve the explained variance of resultant equations.
In addition, analysis of data separately by sex implicitly states that males and females have a different estimate of variation. Incorporating sex as a variable also results in smaller standard errors, since the sample size is that for both groups combined. Despite postulating that the inclusion of sex might require age/sex or height/sex interaction terms to be included, the addition of these added little to the fit of any individual model.
A further methodological issue relates to the technique used for measuring static lung volumes. Whilst the most recent ERS/ATS guidelines on measurement of lung volumes by body plethysmography31 recommend a ‘panting’ technique, this method was not available using our equipment at the start of the study (Masterlab 4.5 Erich-Jaegar, Wurzberg, Germany). However in a comparison of the two techniques, Roca et al8 found no significant difference in results for FRC obtained using either panting or tidal breathing methods.
Spirometry and static lung volumes—In keeping with a previous study incorporating sex into a single equation,32 our equations for FEV1 and FVC had high values for R2 with 80% of population variability being explained by the equation. This compares favourably with typical R2 values for spirometric data of the order of 16–85%.17,26,29 Lower values for R2 are often seen for static lung volumes33—but using our equations, similarly high values of explained variance were seen for TLC and SVC. R2 values higher than previously reported using plethysmography 8 were also found for other static lung volumes.
At the time of their original publication, the reference equations for the ECSC10 represented a major effort in standardisation of pulmonary function testing.16 However, our findings that measured FEV1 values were on average 0.36 L (female) and 0.48 L (male) higher than those predicted by the ECSC equations, with FVC values showing an even greater discrepancy, suggest that the spirometry equations from this source can no longer be considered suitable for contemporary use in New Zealand.
These results are consistent with previous European and Australian findings16,17,30 and are not unexpected since the ECSC equations were compiled from a series of studies conducted between 1954–1980. The under prediction seen with these equations is likely to be due to both anthropomorphic population changes as well as changes in and standardisation of lung function measurement.17
Whilst the equations of NHANES26 have recently been shown to be suitable for use for spirometry in New Zealand34 these do not extend to cover the measurement of static lung volumes or gas transfer. Indeed, fewer reference equations are available for static lung volumes than for spirometry.4
We found that the European equations,10,11 which were mostly derived from studies using gas dilution methods, under-predicted measured values for FRC and TLC in our cohort.
In contrast, a direct comparison between our equations and those of Roca et al8 showed much smaller differences - these equations were also produced using a plethysmographic technique. As a result, it is likely that the differences between the European and WRS equations for static lung volumes can be attributed to both measurement techniques and historical anthropomorphic changes. This means that that the European equations are probably no longer suitable for predicting static lung volumes in contemporary New Zealand subjects of European origin, particularly since in New Zealand such measurements are typically made using plethysmography.
Newer spirometric variables—The measurement of FEV6 has recently been recognised as an important addition in the spirometric assessment of airflow obstruction18,19 and, with FEV1/FEV6,35 offers particular advantages in screening for and assessment of COPD.
Since these are ‘new’ parameters, fewer reference equations are available for their prediction.26,29 Our results provide further reference values applicable across a wide age range.
Gas transfer—Traditionally it has been recognised that measurement of DLCO between different laboratories is subject to wide variation.24 The results for DLCO from our cohort show good agreement with the predicted values derived using the European equations of the ECSC/ERS.10,12 The explained variance of 71% for our new equation for DLCO was significantly higher than that reported in other studies36–38 possibly due to the incorporation of sex into the prediction equation.
We have presented a complete set of reference equations for a comprehensive range of lung function measurements suitable for use in an adult New Zealand population of European origin.
The development of comparable reference equations for Maori and Pacific Island people is a priority.
Author information: Suzanne Marsh, Medical Research Fellow, Medical Research Institute of New Zealand (MRINZ); Sarah Aldington, Medical Research Fellow, MRINZ; Mathew Williams, Respiratory Scientist, MRINZ; Mark Weatherall, Medical Consultant and Biostatistician, Wellington School of Medicine and Health Sciences; Philippa Shirtcliffe; COPD Programme Director, MRINZ; Amanda McNaughton, Senior Research Fellow, MRINZ; Alison Pritchard; IT Manager, MRINZ; Richard Beasley, Director, MRINZ; Wellington
Acknowledgements: This study was supported by a research grant from GlaxoSmithKline. We thank Joan Soriano and Harvey Coxson for their helpful comments in the design of the WRS study and interpretation of the results; P Sherwood Burge for his comments on the manuscript; Denise Fabian for her help with the Wellington Respiratory Health Survey and in producing the manuscript; and Avrille Holt, Patricia Heuser, and Eleanore Chambers for their help in conducting the questionnaires.
Correspondence: Richard Beasley, Medical Research Institute of New Zealand, PO Box 10055, Wellington. Fax: (04) 472 9224; email: Richard.Beasley@mrinz.ac.nz
This article was corrected on 18 May 2007 to reflect the Erratum at http://www.nzma.org.nz/journal/120-1254/2551
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals