Reference values for spirometry in elderly individuals: a cross-sectional study of different reference equations

Spirometry is the single most important test for the evaluation of respiratory function and screening of general respiratory health [1]. Spirometry interpretation compares measured data with previously obtained predicted values, preferably obtained from a reference population [2]. The predicted values, not only for spirometry but also for other lung function tests, vary with age, sex, standing height, and ethnic group, and are obtained using reference equations [2]. The selection of adequate reference equations for spirometry has been discussed previously [3] and was updated in 2012 with the publication of all-age multi-ethnic reference equations for spirometry (the Global Lung Initiative [GLI]) [4]. These equations included a larger population, more ethnicities, and ages ranging from 3 to 95 years. Swanney and Miller [5], in their editorial for the European Respiratory Journal, emphasized the statistical approach of the GLI, which included a larger age range, as young and elderly subjects show greater variability in predicted values than middle-aged subjects. Therefore, the result was the publication of reference equations for a wide age range that attempted to rectify the relative lack of reference data for elderly subjects [6, 7]. Before the GLI publication, specific studies of elderly populations showed substantial variability in predicted values, which may lead to interpretation inaccuracies [7‐10] that become pertinent in elderly populations as they continue to be studied. The adoption of new GLI equations and the possible effect on lung function test interpretation had already been studied in adults aged less than 85 years [11, 12], children, and adolescents [13]. To our knowledge, there are no significant data available regarding the effects of spirometry interpretation in an extremely aged population.

The reference equations should be derived from a population with the same characteristics as the population of the tested individuals; at least 53 studies on this subject were published between 1995 and 2004 [14]. In most lung function laboratories (LFLs), the choice of reference values varies; recommendations from the National Health and Nutrition Assessment Survey (NHANES III) [15] are commonly used in the United States of America (USA), and those from the European Community of Steel and Coal (ECSC) [16] are commonly used in Europe. The population ages and ethnic characteristics are the primary differences between the GLI, NHANES III, and ECSC data. The NHANES III equations are derived from a population sample of Caucasian, Mexican, and African-American descent between 18 and 80 years of age, while the ECSC equations are derived from a Caucasian sample between 18 and 70 years of age. The GLI was a task force endorsed by five international societies and used a multi-ethnic sample from 3 to 95 years of age.

As each recommendation used a different population, it is clear that the reference equations obtained must also be different, creating potential problems in result interpretation for laboratories, clinicians, and technologists. Thus, it is important to identify the implications of adopting different equations. According to Brazzale et al. [17], changing to the GLI equations could affect the interpretation of spirometry values, as their direction and magnitude are dependent on which reference data are used in practice.

The aim of this study was to compare spirometry results expressed as percentages of the predicted values and spirometry interpretation using ECSC, NHANES III, and GLI reference equations in elderly subjects.

Out of the 817 elderly subjects invited to participate in spirometry testing, only 307 accepted. However, 45 elderly subjects were excluded because they did not produce a quality spirometry result. Moreover, as the predicted values obtained for Caucasians are consistently different from those obtained for non-Caucasians [19, 20], two elderly subjects who were not Caucasian were also excluded from the analysis. A total of 260 elderly subjects were included in this study. Subjects who underwent spirometry had a mean age of 83.0 ± 6.46 years, and 181 (69.6%) were women.

In this study we assessed differences between values obtained from distinct sets of reference equations (ECSC, NHANES III, and GLI) in a sample of elderly subjects recruited at ECC. Moreover, the LLN was used for spirometry interpretation to discuss possible disparities. In fact, some differences were expected, as each reference equation was obtained by applying different statistical models to distinct populations.

Our findings indicate that higher values are calculated for %FVC and %FEV₁ when the ECSC reference equations are used, allowing us to conclude that these equations provide lower predicted values. Higher predicted values for FVC and FEV₁ are obtained with GLI equations, as our results show lower %FVC and %FEV₁ values using these reference equations. However, %FEV₁/FVC is underestimated using ECSC equations and overestimated using NHANES III equations. As LLN is obtained by considering predicted values, the highest LLNs are found for FVC and FEV₁ using GLI equations; for FEV₁/FVC, the highest LLNs are found using ECSC equations. These results suggest different rates of ventilatory disturbances based on spirometry. Since the LLN for FEV₁/FVC differs by 10% between the GLI and ECSC equations and by 9% between the NHANES III and ECSC equations, the impact on the rate of airway obstructions was substancial when changing from ECSC equations. In our sample, ECSC reference equations identified more subjects with airway obstruction (52.3%) than either the NHANES III or ECSC reference equations. The highest LLN for FVC establishes a stricter criterion for a spirometric restrictive pattern. The GLI equation results in a higher mean LLN for FVC (100 mL and 400 mL compared with NHANES III and ECSC, respectively), and for that reason, the percentage of elderly subjects with restrictive defects was substancial (29.6%). The level of agreement was poor-to-slight between ventilatory defect proportions using the GLI equations as a reference compared with LLN criteria from ECSC equations and tended to be lower in more elderly subjects. Another relevant finding was that all differences observed were more evident in women and in subjects who were more elderly.

Previous studies were conducted to explore the applicability of GLI equations to different populations. Overall, the study populations were heterogeneous and comparisons with current data may be compromised. However, some considerations can be made.

We found discrepancies in airway obstruction and the spirometric restrictive pattern rate in most previous studies. Our data reflect more respiratory disturbances based on spirometry than in any other study. This finding is not surprising, as it is well known that lung volume decline is age-dependent [21‐23]. Moreover, our sample included institutionalized elderly subjects; these individuals are potentially more dependent and have chronic diseases such as respiratory diseases. The prevalence of airway obstruction was markedly higher in only one study [12]. The authors calculated the predicted values for spirometry parameters using the same reference equations used in the present study. However, despite the young sample age, the data were from patients referred to pulmonary laboratories of tertiary hospitals, so it is probable that the prevalence of respiratory disease was higher.

Previous evidence showed both similarities and discrepancies when comparing respiratory disturbance rates between the three sets of reference equations. We found more airway obstructions using ECSC equations while more spirometric restrictive patterns were observed using GLI equations. Globally, previous studies have found that adopting the GLI reference equations had a minor impact on the rate of airway obstruction rather than on the rate of spirometric restrictive patterns where the impact was higher [12, 13, 17]. A French study [24] determined that GLI equations increased the rate of airway obstruction by 2.2% and the rate of spirometric restrictive patterns by 5.8% compared with ECSC equations. Between GLI and NHANES III equations, NHANES III equations identified 3% more airway obstruction in male subjects and 2.3% more obstruction in female subjects [25]. The primary differences between the referenced studies and ours are the age ranges and the older mean age, that was higher in the present study.

The lowest LLN for FVC and FEV₁ using ECSC equations was found in other studies [17], although our study showed a higher FEV₁/FVC ratio using ECSC equations, which contradicts previous studies [12, 17]. To justify this finding, we suspect that decreasing FVC and FEV₁ values do not follow the same proportions, so the ratio does not show the same trend.

More recent studies were conducted to test the applicability of the GLI reference equations in comparison with those commonly used in the respective study populations [25‐28]. A group of 1000 healthy, non-smoking, native Finnish subjects aged 18–83 years were evaluated and it was found that GLI equations underestimated lung volumes, and especially the FVC [26], which was a finding that contradicted our data. In this study, the authors compared the GLI reference equations with their own equations, and despite the lack of comparative studies, it has been suggested that Northern populations may have slightly larger lung volumes, a fact that can justify this disparity [26]. The mean %FEV₁ predicted values derived from ECSC equations showed the lowest values in both female and male subjects in a Netherlands study [27], which are findings that contradict ours. An important difference between the two studies is that the mean age was much lower in the previous study than in ours. However, in a sample of elderly Brazilian subjects with a mean age approximately equal to that in our study, the LLN was similar to that in our study when using GLI equations [28]. Linares-Perdomo et al. [25] found that differences between NHANES III and GLI equations are generally small and probably not clinically important for most patients. However, the authors suggest caution when selecting prediction equations for elderly patients, and especially for tall and short patients. In fact, our sample is characterized by a significant percentage of elderly women patients (69.6%) with a mean height of 1.56 m.

The ability of elderly individuals to achieve a good spirometry performance deserves further discussion. To our knowledge, the first study concerning the quality control of spirometry performance in elderly individuals was performed in 2000. The primary results showed that the sources of variability in performance are numerous, and include motor and sensory deficits, dementia, depression, and malnutrition [29]. Despite the fact that more time was required to complete the examination to obtain the necessary quality standards, in this study, all spirometry tests adhered to international guidelines for acceptability [1] and repeatability [18]. According to Pezzoli et al. [30], age cannot be considered a risk factor for a poor spirometry performance, as in their study the majority of elderly patients with no cognitive or functional impairment underwent spirometry tests according to international guidelines.

The study subjects ranged in age from 65 to 95 years. ECSC reference equations were obtained from a sample of subjects of European descent with ages that ranged from 18 to 70 years. In all, 96.2% of the elderly subjects included in our study were more than 70 years old. NHANES III included a random sample of the US population with an age range of 8–80 years living in households. It provides regression equations for predicted values and LLNs for the three primary racial/ethnic groups in that country: Caucasians, African-Americans, and Mexican-Americans. The applicability of NHANES III data to the present sample is associated with two major limitations. First, 76.2% of the elderly subjects in the previous study were more than 80 years old, and second, the U.S. population has different anthropometric data characteristics than those found in Europe. The predicted values for subjects older than the maximum ages evaluated by the NHANES III and ECSC were extrapolated beyond 80 and 70 years, respectively. This fact may have caused bias in the results; however, these reference equations are still used in some LFLs and in community-based spirometry studies, so it is necessary to test their applicability to elderly subjects.

The spirometric restrictive pattern should be interpreted carefully. In fact, the measurement of FVC by spirometry is useful to exclude restriction, but there are limitations to the identification of disturbances [31, 32]. An accurate restrictive pattern can only be identified by measuring the total lung capacity (TLC).

An ageing population represents the reality in most developed countries, and in recent years, a set of initiatives has been developed for active ageing that includes more opportunities for health and social participation to enhance the quality of life in the elderly population. Ageing is also accompanied by a decline in lung function, and this population is more susceptible to acute and chronic pulmonary diseases. We believe that the main conclusions of this study determined two goals for further investigations. One is the requirement to continue the validation of GLI equations in elderly populations, assessing discrepancies in the results between men and women. The GLI’s new update is the first approach that allows calculations of the predicted, lower, and upper limits of normality for subjects up to 95 years old. Another advantage of the GLI equations is that z-scores can be calculated, allowing clinicians to interpret lung function results independent of age, height, sex, and ethnic group, which decreases the possible bias of using predicted values. Therefore, testing the benefit of using z-scores in an elderly population will update lung function procedures, as this practice is only used in paediatric populations.

The purpose of this study was not to conclude which reference equation should be used for elderly subjects but rather to demonstrate the impact of each reference equation on the percentages of predicted values and proportions of ventilatory disturbances. Moreover, it is important to show how the clinical intervention and patient’s follow up may be compromised by the reference equation used. In fact, it is possible to conclude from this study that changing from ECSC to GLI equations will have a meaningful impact on classifications with fewer obstructed and more restricted subjects. The data in the present study showed meaningful differences in spirometric outcomes, and consequently, in the proportion of ventilatory disturbances.