The aim of this study was to describe spirometric reference equations for healthy never-smoking European adults aged 65–85 yrs and to compare the predicted values of this sample with those from other studies including middle-aged and/or older adults.
Reference equations and normal ranges for forced expiratory volume in one second (FEV1), forced vital capacity (FVC), forced expiratory volume in six seconds (FEV6), FEV1/FVC ratio and FEV1/FEV6 ratio were derived from a healthy subgroup of 458 subjects aged 65–85 yrs. Spirometry examinations followed the 1994 American Thoracic Society recommendations and the quality of the data was continuously monitored and maintained. Reference values and lower limits of normal were derived using a piecewise polynomial model with age and height as predictors.
The reference values of FEV1 and FVC from the present study were higher than those given by prediction equations from the European Community for Coal and Steel. By contrast, use of prediction equations from Caucasian-American elderly subjects (Cardiovascular Health Study) consistently overpredicted FVC and FEV1 in females by 8.5 and 2.1%, respectively. In males, equations from the Cardiovascular Health Study overpredicted FVC by 2.8%, whilst underpredicting FEV1 by 2.5%.
In conclusion, these results underscore the importance of using prediction equations appropriate to the origin, age and height characteristics of the subjects being studied.
This study was supported by FIS (99/0218) and NEUMOMADRID grants.
Spirometry is probably the most important tool used in screening for pulmonary disease and is the most frequently performed pulmonary function test. Although the average age of patients tested at pulmonary function laboratories each year is ∼60 yrs old, many of the reference equations commonly used for the prediction of normal spirometric values in North America and Europe have been derived from studies that included relatively small numbers of individuals >65 yrs old 1–9. In fact, predicted values for older individuals are often based upon few observations or extrapolations from data acquired in studies of younger adults. However, the application of prediction equations derived from primarily younger adult populations to older adults may be inappropriate because the relationship between lung function, age and height may change with age. In fact, the current international guidelines recommend that spirometry reference equations should, in general, not be extrapolated for ages or heights beyond those covered by the data that generated them 10, 11.
Valid reference values for spirometric parameters in healthy elderly Afro-Americans 12 and Japanese-American elderly males 13 have been previously reported. Only two sets of standards have been published on lifetime nonsmokers in Caucasian-American elderly subjects 14, 15. Although significant differences between Caucasians of American and European origin have been suggested 16, no study has collected pulmonary measurements for both sexes across a large sample of elderly European subjects. Enright et al. 14 derived spirometric prediction equations from a reference population of healthy individuals aged 65–85 yrs. However, the study did not provide reference equations for spirometry variables other than forced vital capacity (FVC), forced expiratory volume in one second (FEV1) and FEV1/FVC ratio 14. Although with some exceptions 9, 17, many previous prediction equations for elderly subjects were linear 12–14 and, therefore, did not reflect accelerating decline with age. Finally, only one previous study 6 provides reference values for elderly subjects for forced expiratory volume in six seconds (FEV6) and FEV1/FEV6 ratio, an acceptable surrogate for FVC for the spirometric diagnosis of obstruction 18.
The purpose of the current study is to describe spirometric reference equations for a cohort of healthy never-smoking Caucasian-European adults aged 65–85 yrs and to compare the predicted values of this sample with those from other studies, including middle-aged and/or older adults.
Materials and methods
The total target population consisted of 466,958 inhabitants aged 65–85 yrs, included in the census register of the Madrid metropolitan area, Spain (760 m above sea level). A random sample of 1,300 subjects proportionally stratified by sex and age (65–69, 70–74, 75–80 and 81–85 yrs) was drawn by electronic selection to approximate the total population distribution.
Eligible persons were invited to participate if they were lifetime never-smokers and had no known history of respiratory or cardiovascular disease. Over a 12-month period, starting in February 2001, potential participants were sent an explanatory letter, interviewed by telephone to determine eligibility and, then, scheduled for the baseline clinical examination. During this time, the study was explained using local mass communication media (radio and television) to increase the acceptation rate. Among those contacted, 46.5% were ineligible and 16.3% of those eligible refused to participate.
Clinical evaluation was based on an extended combination of the European Community for Coal and Steel questionnaire on respiratory symptoms 19, a physical examination, complete blood count and blood chemistry, a conventional chest radiograph evaluation and 12-lead resting electrocardiography (ECG).
The exclusion criteria were: history of chest injuries; exposure to substances known to cause lung injury; respiratory disease (self-reported or medical doctor-diagnosed asthma, pulmonary tuberculosis, pneumonia, frequent bronchitis, emphysema or chronic bronchitis); respiratory symptoms during the last 12 months (dyspnoea, chronic cough, wheezing or phlegm); hypertension or hypotension; clinically relevant alterations of the physical examination of the heart, lungs and chest wall; abnormal chest radiographs; major ECG abnormalities; pitting ankle oedema; diabetes (self-reported or fasted glucose level >140 mg·dL−1); and the use of diuretics, cardiac glycosides or β-adrenergic blocking agents.
The study was approved by the local Ethics Committee. Informed consent was obtained from all subjects.
All tests were performed by a single technician experienced in lung function testing (A. Dorgham). Spirometry was recorded with a pneumotachograph (MasterLab 4.6; Jaeger, Wurtzburg, Germany). The system was calibrated with a 3-L syringe each morning and recalibrated ≥3–4 h. The technician also performed a daily biological control by assessing his own lung function. Standing height was measured to the nearest 0.5 cm without shoes, with the subject's back to a vertical backboard. Both heels were placed together, touching the base of the vertical board. Subjects were weighed whilst wearing indoor clothing without shoes, and body mass index (BMI=weight/height2; expressed in kg·m−2) and body surface area (BSA=0.20247×height0.725×weight0.425; expressed in m2) were calculated. Age was recorded to the nearest birthday. Barometric pressure, temperature and relative humidity were registered every morning, and the integrated volumes were automatically converted from ambient temperature and pressure, (saturated with water vapour) to body temperature and ambient pressure (saturated with water vapour) conditions.
Spirometry flow/volume loops were conducted in accordance with American Thoracic Society (ATS) recommendations 11. At least three acceptable trials were required, defined as a good start of test (extrapolated volume of <5% ofFVC or 0.15 L, whichever was larger), at least 6 s of expiration and a plateau in the volume/time curve (change in volume <30 mL for ≥2 s). Time zero of each manoeuvre used the back-extrapolation technique 20. As recommended by the ATS, data that did not meet reproducibility criteria were not excluded, but subjects were asked to perform up to a maximum of eight manoeuvres in an attempt to obtain reproducible results 11. The highest FEV1, FEV6, and FVC from tests of acceptable quality were used for analysis. The other parameters were taken from the trial with the largest sum of FVC and FEV1.
For the measurement of forced inspiratory volumes, patients exhaled slowly from tidal breathing until the residual volume was achieved, with subsequent forceful inspiration until total lung capacity was reached. At least three measurements of forced inspiratory volumes were taken. In analogy with ATS criteria 10, 11, from two acceptable manoeuvres (difference <5%), the highest value of the forced inspiratory volume in one second (FIV1) was chosen for analysis. Peak inspiratory flow rate and the forced inspiratory flow when 50% of the vital capacity has been inhaled were taken from the test with the largest sum of FVC and FIV1.
Independent variables considered for inclusion in the models were as follows: age, age2, age3, standing height, height2, weight, weight2, BMI, BMI2, and BSA. The effect of logarithmic and square root transformations of pulmonary function parameters prior to modelling was also examined.
In the multiple linear regression analysis, predictor variables were retained only if their addition significantly improved (p<0.05) the fraction of explained variability. Other aspects explored included residual standard deviation (rsd), changes in the distribution of the residuals and the homogeneity of the variance over the predictors. Statistical significance was assumed for p<0.05. The assumptions of linearity and distributional normality were controlled. Residuals values were plotted against age and height to examine for heteroscedasticity. The lower limit of normal (LLN) range was calculated as follows: 1The selection of prediction equations for comparison was based on common use 2–9 and inclusion of the elderly 13–15. Differences between observed values and values predicted by the prediction equations are given as mean difference in per cent of mean observed values, mean squared difference and standardised prediction deviation (i.e. mean prediction deviation/rsd of the corresponding prediction equation). For comparisons among different authors, LLN was calculated using the rsd of the corresponding equation. The differences between predicted values based on the prediction equations from the present study and others are given as Bland and Altman plots.
A total of 583 subjects underwent clinical evaluation. In total, 76 subjects were excluded by dyspnoea (n=24), cough (n=17), wheezing (n=13) and for several previously unknown diseases, such as chronic obstructive pulmonary disease (n=11), asthma (n=8) or scoliosis (n=3). Of the 507 subjects (314 females and 193 males) who were entered into the study, technically acceptable tests were found in 458 (279 females and 179 males). A total of 49 subjects (7.2% of males and 11.1% of females) were excluded from analysis because the expiration time was <6 s (n=37) or because of a poor test start (n=12). The elderly persons who were excluded or were ineligible for the study were similar in age, height and weight to those who were included.
The age distribution of the females and males in the analysed sample (table 1⇓) demonstrates adequate representation of the study population. Details of the anthropometric and spirometric data in both sexes are shown in table 2⇓. No significant differences in these parameters were found between excluded subjects and the analysed sample.
The spirometry reference equations from the healthy elderly European females and males are given in tables 3 and 4⇓⇓. It was not found that the addition of transformations significantly improved the predictability of the regression equations. Preliminary multiple regression analysis indicated that neither BMI nor BMI2 were associated with FVC, FEV1 or any other spirometric parameter in either sex. No significant interaction was found between age and height.
Analysis of residuals showed that homoscedasticity was present in all equations. Regression analysis of these residuals showed neither statistically significant slopes nor correlation coefficients. The residuals corresponding to these models didnot differ significantly from a Gaussian distribution in all spirometric parameters, as determined by the Shapiro-Wilk test. Therefore, one-sided lower 95% prediction intervals were used to determine the LLN lung functions 4, 21.
Table 5⇓ shows the differences between the observed spirometric values found in the subjects of the current study and the values calculated from several prediction equations. Aside from the current authors' equations, the closest agreements for FVC were with Crapo et al. 3, Langhammer etal. 9, Hankinson et al. 6 and Knudson et al. 2 in females, and with Hankinson et al. 6, Langhammer et al. 9, McDonnell et al. 15 and Enright et al. 14 in males. In males, the closest agreements for FEV1 were with Paoletti et al. 7, Hankinson et al. 6, Enright et al. 14 and the European Community for Steel and Coal 4. Meanwhile, in females, the closest FEV1 agreements were with Enright et al. 14, Roca et al. 5, Hankinson et al. 6 and Crapo et al. 3.
To compare the current authors' reference equations with other prediction equations, the difference in predicted FEV1 (present study equation–each other equation) by the mean predicted FEV1 are illustrated in figures 1 and 2⇓⇓ for females and males, respectively. In females, a proportional increase for FEV1 with respect to Knudson et al. 2, Paoletti et al. 7 and Brändli et al. 8 was found (fig. 1⇓). In males, the relationship increased proportionally when the present prediction values for FEV1 were compared with those from Brändli et al. 8 and Sharp et al. 13, whilst it decreased proportionally with respect to Knudson et al. 2 (fig. 2⇓). In contrast, proportional decreases of relationship with respect to Hankinson et al. 6 were found for FEV1 in both sexes.
For both females and males, European Community for Steel and Coal equations underpredicted FEV1. In contrast, Crapo et al. 3, Roca et al. 5, Brändli et al. 8 and Langhammer et al. 9 overestimated both FVC and FEV1 in males.
The current study provides equations for predicting lung function values in a population of healthy older European adults. The results confirm that reference equations should not be extrapolated, in general, for ages or heights beyond those covered by the data that generated them. For patients >65 yrs of age, the current study showed that the most commonly used sets of reference equations may lead to inaccurate interpretations.
The reference values for FEV6 provided in this study are not widely available in the literature. To the current authors' knowledge, only the previous study by Hankinson et al. 6 published results for the FEV6, and included 90 males and 236 females in the 66–80 yrs age range. In contrast to the scarce reference equations for FEV6, this parameter could be a potential surrogate for FVC in those situations where long exhalation times are impractical or unwarranted, particularly in elderly or severely obstructed subjects. Recently, it has been suggested that FEV1/FEV6 could predict lung function decline in adult smokers 22.
The current results can be contrasted with the Japanese-American and American males in the Honolulu Heart Program (HHP) 13 and the Cardiovascular Health Study (CHS) 14 cohorts, respectively. The height and age characteristics of the currently studied males are in between the HHP and CHS males, probably reflecting contrasting design characteristics. In contrast to HHP and CHS cohorts, the current study subjects were age stratified according topopulation characteristics. Consequently, the percentage of subjects >75 yrs old was lower than in the HPP study. Most notably, the HHP males are much thinner than the CHSand the currently studied males, as indicated by BMI distributions.
The exact definition of a “healthy” group is difficult to agree upon 14, 15, 23. Previous studies have used many different criteria. The ATS spirometry interpretation workshop only states that subjects should be “never-smokers, free of respiratory symptoms and disease” 10. In contrast to Enright et al. 14, who did not exclude previous smokers with <5 pack-yr history of smoking who had quit smoking >5 yrs previously, all of the current study patients were lifelong nonsmokers.
The FVC and FEV1 age regression coefficients for the current female and male groups were similar to those of Enright et al. 14 and McDonnell et al. 15. Moreover, the age coefficients for FEV1 in elderly subjects were nearly identical to those reported from younger cohorts of healthy persons (−32 and −44 mL annual change in FEV1 in females and males, respectively). Several longitudinal studies suggest a small degree of nonlinearity in the downward slope 2, 24. Probably as a result of this, the addition of a nonlinear age term improves the strength of the current authors' regression equations, despite the narrower age range of the current healthy group.
In contrast, the current data suggest that FVC in males >65 yrs old have a stronger, more negative relationship with age and a weaker positive relationship with height than do individuals <65 yrs old. The current authors' observation that FVC was related to age3 is consistent with observations in several longitudinal studies that loss of lung function may be accelerated in the elderly 24, 25. However, as is clearly shown in figure 3⇓, the net result of the different evolution of FVC and FEV1 is a less declining FEV1/FVC with ageing. Premature ending of the spirometric manoeuvre could explain the differences in the age coefficients for FVC and FEV1; however, all the subjects included in the current study reached the required expiratory flow plateau and, moreover, no relationship between expiratory time and FEV1/FVC coefficient was detected. Recently, Pezzoli et al. 26 have reported that 81.8% of elderly subjects with respiratory symptoms were able to perform spirometry according to international guidelines. It seems reasonable to assume that, in elderly patients with no respiratory symptoms, the percentage of satisfactory manoeuvres would be at least similar and that premature ending would be infrequent.
Studies of middle-aged adults have demonstrated that both extremes of body weight are associated with lower FVC 24, 27, 28. In the current study, 43 females had a weight that was 20 kg below the mean; however, no significant differences in FVC, FEV1 or FEV6 were found between females with a weight that was 20 kg below the mean weight and females with average body weight.
In the current male subjects, no spirometric parameter was related to weight; however, in females, FVC, FEV1 and FEV6 were related to weight and BSA. In middle-aged subjects, BMI has recently been considered to be an additional independent variable in models for deriving spirometric prediction equations 17, 29. However, in the current study, no relationship between spirometric parameters and BMI was detected. It is possible that the narrow weight range of the studied patients explains the absence of this relationship. Nevertheless, it should also be considered that FVC and FEV1 depend more on body composition than BMI, especially in males 30. Therefore, the lack of discrimination in BMI of the changes in both fat and muscle experienced by elderly subjects could explain its uselessness to reference equations for spirometry in these subjects.
The FEV1/FVC ratio is generally used as a sensitive index to separate patients with borderline to mild airflow limitation from those with normal spirometry. A general rule often used by clinicians with middle-aged patients is that values <70% indicate obstruction. However, large cross-sectional and longitudinal studies of healthy middle-aged adults show that this ratio declines with age. The current age-related change in predicted FEV1/FVC ratios matches the other selected studies (fig. 3⇓). With increasing age at constant height, the present study predicted that FEV1/FVC values decline from 80 to 70% in females and from 79 to 73% in males, with lower limits of normal from 71 to 68% and from 70 to 64%, respectively. In this sense, Hardie et al. 31 have recently suggested that the criteria used to determine the normal limits of FEV1/FVC need to be age specific. The current results demonstrated that the old rule of thumb that 70% is the lower limit of the normal range should not be used with elderly patients; otherwise, many false-positive interpretations for airway obstruction will result. The FEV1/FEV6 ratio did not allow the current authors to obviate this problem because it is also dependent on age.
As prediction equations derived from cross-sectional data are primarily used as a screening tool to identify individuals with lung function below the expected range, the utility of any particular reference equation depends upon its ability to correctly identify individuals with lung function below the lower limit of normal. Some authors have defined the LLN as that value above which the results of 95% of the normal population lie, working under the assumption that larger values have larger variances. However, if skewed distributions are transformed to normalise their shape, the subtraction of 1.645 sd may still be used to estimate the LLN. In males, the equations for the LLN from Sharp et al. 13, Enright et al. 14 and McDonnell et al. 15 identify none of the current study's participants as being below the FEV1 LLN. In contrast, Enright et al. 14 and McDonnell et al. 15 identified 3.2 and 8.2% of females in the current study as being below the FEV1 LLN. Whilst some of the differences among these studies in predicting mean and LLN in the elderly may be due to inclusion of different age ranges, the effects of dissimilarity in underlying populations, measurement methods and reference group exclusion criteria also play a large role.
Table 6⇓ shows the main characteristics of the studies that have provided the reference equations with which the current authors compared theirs. It is evident that the differences in age range, body-mass range or selection criteria of the sample could explain some of the differences obtained. The type of analysis used seems to be less relevant, given that the nonlinear equations of Langhammer et al. 9 do not better adjust to the current sample than the linear equations, especially for FVC.
Some of the differences observed could also reflect the choice of instrument. Instrument differences in the measurement of FVC could, theoretically, be as large as 7% and still meet ATS standards, which only require them to be within ±5.5% of the target values. As such, two aspects of the current study could be worthy of special consideration: 1) the larger differences occurred in FVC and not in FEV1; and, moreover, 2) the mean differences from the Hankinson et al. equations 6 were smaller for FEV6 than FVC. Both circumstances could be related to the influence of instrumentation on exhalation time. Minimal differences in the instruments used for measuring could be exaggerated by the greater variability at the end of the forced expiratory manoeuvre. Likewise, and since the spirometer used by Hankinson et al. 6 was a rolling-seal spirometer, it cannot be ruled out that the effect of cooling with longer expiratory times could have affected the results.
In conclusion, the current authors have developed reference equations for the prediction of lung function of older adults. Differences among studies in predictions of lung function or in identification of individuals with lung function values below the lower limit of normal may be due to differences in the age range of the reference subjects, but are also likely to be contributed to by differences in exclusion criteria, different measurement methods and other differences in the underlying populations. These results underscore the importance of using prediction equations appropriate to the ethnicity, age and height characteristics of the population to whom inferences are to be applied.
The current authors acknowledge the excellent technical assistance provided by A. Alvarez, P. Librán, A. Pérez and C. Suárez.
↵For editorial comments see page 341.
- Received July 30, 2003.
- Accepted April 19, 2004.
- © ERS Journals Ltd