Interpretative strategies for lung function tests

R. Pellegrino, G. Viegi, V. Brusasco, R. O. Crapo, F. Burgos, R. Casaburi, A. Coates, C. P. M. van der Grinten, P. Gustafsson, J. Hankinson, R. Jensen, D. C. Johnson, N. MacIntyre, R. McKay, M. R. Miller, D. Navajas, O. F. Pedersen, J. Wanger


Edited by V. Brusasco, R. Crapo and G. Viegi

Number 5 in this Series


This section is written to provide guidance in interpreting pulmonary function tests (PFTs) to medical directors of hospital-based laboratories that perform PFTs, and physicians who are responsible for interpreting the results of PFTs most commonly ordered for clinical purposes. Specifically, this section addresses the interpretation of spirometry, bronchodilator response, carbon monoxide diffusing capacity (DL,CO) and lung volumes.

The sources of variation in lung function testing and technical aspects of spirometry, lung volume measurements and DL,CO measurement have been considered in other documents published in this series of Task Force reports 14 and in the American Thoracic Society (ATS) interpretative strategies document 5.

An interpretation begins with a review and comment on test quality. Tests that are less than optimal may still contain useful information, but interpreters should identify the problems and the direction and magnitude of the potential errors. Omitting the quality review and relying only on numerical results for clinical decision making is a common mistake, which is more easily made by those who are dependent upon computer interpretations.

Once quality has been assured, the next steps involve a series of comparisons 6 that include comparisons of test results with reference values based on healthy subjects 5, comparisons with known disease or abnormal physiological patterns (i.e. obstruction and restriction), and comparisons with self, a rather formal term for evaluating change in an individual patient. A final step in the lung function report is to answer the clinical question that prompted the test.

Poor choices made during these preparatory steps increase the risk of misclassification, i.e. a falsely negative or falsely positive interpretation for a lung function abnormality or a change in lung function. Patients whose results are near the thresholds of abnormality are at a greatest risk of misclassification.


General issues

Interpretation of PFTs is usually based on comparisons of data measured in an individual patient or subject with reference (predicted) values based on healthy subjects. Predicted values should be obtained from studies of “normal” or “healthy” subjects with the same anthropometric (e.g. sex, age and height) and, where relevant, ethnic characteristics of the patient being tested. Ideally, reference values are calculated with equations derived from measurements observed in a representative sample of healthy subjects in a general population. Reference equations can also be derived from large groups of volunteers, provided that criteria for normal selection and proper distribution of anthropometric characteristics are satisfied. Criteria to define subjects as “normal” or healthy have been discussed in previous ATS and European Respiratory Society (ERS) statements 5, 7, 8.

Height and weight should be measured for each patient at the time of testing; technicians should not rely on stated height or weight. Height should be measured with a stadiometer, with shoes off, using standard techniques (patient standing erect with the head in the Frankfort horizontal plane) 9. When height cannot be measured, options include using stated height or estimating height from arm span, as indicated in a previous document from this series and other publications 1, 10, 11.

Specific recommendations for selecting reference values to be used in any lung function laboratory have also been discussed 3. These include the following: matching age-range, anthropometric, race/ethnic, socio-economic and environmental characteristics between subjects investigated by the laboratory and the reference population from which the prediction equations have been drawn; using similar instruments and lung function protocols in the reference population as in the laboratory; and using reference values derived by valid and biologically meaningful statistical models, taking into account the dependence of lung function with age. If possible, all parameters should be taken from the same reference source. For example, forced vital capacity (FVC), forced expiratory volume in one second (FEV1), and FEV1/FVC should come from the same reference source.

The subjects being tested should be asked to identify their own race/ethnic group, and race/ethnic-specific reference equations should be used whenever possible. If such equations are not available or are unsuitable for a particular setting, a race/ethnic adjustment factor based on published data may be used for lung volumes. The use of adjustment factors is not as good as specific race/ethnic equations 12. An example of adjustment factors is the finding that populations using standing height as the measure of size tend to overpredict values measured in Black subjects by ∼12% for total lung capacity (TLC), FEV1 and FVC, and by ∼7% for functional residual capacity (FRC) and residual volume (RV) 5. A race/ethnic adjustment factor of 0.94 is also recommended for Asian Americans based on two recent publications 13, 14. Such adjustment factors should not be applied to the FEV1/FVC or FEV1/vital capacity (VC) ratios. The use of sitting height does not completely account for race/ethnic differences in pulmonary function 15. If a race adjustment factor is used, a statement should be included in the report, along with the race adjustment value used.

Differences in the evaluation of lung function using different sets of reference equations have been documented 16, 17. Ideally, spirometric reference values should be derived from a population similar to the individual subject using the same kind of instrument and testing procedure.

There have been recommendations to compare selected reference equations with measurements performed on a representative sample of healthy subjects tested in each laboratory. The reference equation that provides the sum of residuals (observed – predicted computed for each adult subject, or log observed – log predicted for each subject in the paediatric age range) closest to zero will be the most appropriate for that laboratory 7. However, for spirometry, a relatively large number of subjects (i.e. n = 100) is necessary to be confident that a significant difference between the published reference equations and the values from the local community does not exist 18. Therefore, the suggestion is impractical for most laboratories.

When using a set of reference equations, extrapolation beyond the size and age of investigated subjects should be avoided 7. If a patient's age or height is outside the limits of the reference population, a statement in the interpretation should indicate that an extrapolation has been made.

Publications on reference equations should include explicit definitions of the upper and lower limits of the normal range, or provide information to allow the reader to calculate a lower range 5. For each lung function index, values below the 5th percentile of the frequency distribution of values measured in the reference population are considered to be below the expected “normal range” 5. If the reference data have a normal distribution, the lower 5th percentile can be estimated as the 95% confidence interval using Gaussian statistics. If the distribution is skewed, the lower limit should be estimated with a nonparametric technique, such as the 95th percentile. The practice of using 80% predicted as a fixed value for the lower limit of normal may be acceptable in children, but can lead to important errors when interpreting lung function in adults 5. The practice of using 0.70 as a lower limit of the FEV1/FVC ratio results in a significant number of false-positive results in males aged >40 yrs and females >50 yrs 12, as well as in a risk of overdiagnosis of chronic obstructive pulmonary diseases (COPD) in asymptomatic elderly never-smokers 19. This discussion has been focused on the lower limit of the reference range. Upper limits are appropriate where the variable can be either too high or too low. Such variables include TLC, RV/TLC and DL,CO. As equipment and techniques for lung function testing improve, advanced mathematical models to describe lung function data are implemented. Furthermore, the characteristics of the populations of “normal” subjects, with respect to nutrition, health status, environmental conditions and other factors, evolve (a phenomenon also described as “cohort effect”). Consideration should be given to updating reference equations on a regular basis, e.g. every 10 yrs, taking into account the applicability of the newer reference equations and the effect on interpretation of longitudinal patient follow-up.

Manufacturers should also provide software that allows users to easily select among a panel of reference equations. They should also allow easy insertion of new equations. The reference values used should be documented on every pulmonary function report with the first author's last name (or organisation) and the date of publication.


The European Community for Coal and Steel (ECCS) 8, 20 and the ATS 5, 21 have both published comprehensive listings of published reference equations for spirometry. A number of additional studies on lung function reference values, dealing with a variety of ethnic/race groups and age ranges, have been published in the last 10 yrs 12, 14, 17, 22, 23.

Spirometric reference equations are usually derived from cross-sectional studies and are subject to “cohort effect”. Few authors have published longitudinal equations covering ages from childhood to the elderly 2426, and there are few published sets of equations that cover volume and flow indices over a wide range of ages 27, 28. Table 1 includes reference equations published from 1995 to August 2004. The table was created from known equations and a MEDLINE search using the keywords “reference equations” and “spirometry”. Its purpose is to recognise and encourage the continuing interest of worldwide researchers in deriving and using reference equations.

In the USA, ethnically appropriate National Health and Nutrition Examination Survey (NHANES) III reference equations are recommended for those aged 8–80 yrs 12. For children aged <8 yrs, the equations of Wang et al. 29 are recommended. Other prediction equations may be used if there are valid reasons for the choice. In Europe, the combined reference equations published in the 1993 ERS statement 8 are often used for people aged 18–70 yrs, with a height range of 155–195 cm in males, and 145–180 cm in females, and those from Quanjer et al. 30 in paediatric ages. Currently, this committe does not recommend any specific set of equations for use in Europe, but suggests the need for a new Europe-wide study to derive updated reference equations for lung function.

Lung volumes

Lung volumes are related to body size, and standing height is the most important correlating variable. In children and adolescents, lung growth appears to lag behind the increase in standing height during the growth spurt, and there is a shift in the relationship between lung volume and height during adolescence 31, 32. Height growth in young males between 12.5 and 18 yrs of age peaks ∼1 yr before the growth rate of weight and FVC, and ∼1.5 yrs before the growth rate of maximum flow at 50% FVC. In young females, growth rates of all spirometric indices decrease over the same age range. Using simple allometric relationships between stature and lung volumes, volume predictions are too high in the youngest age group and too low in the oldest adolescents.

Furthermore, for the same standing height, young males have greater lung function values than young females, and Whites have greater values than Blacks. Lung function increases linearly with age until the adolescent growth spurt at age ∼10 yrs in females and 12 yrs in males. The pulmonary function versus height relationship shifts with age during adolescence. Thus, a single equation or the pulmonary function–height growth chart alone does not completely describe growth during the complex adolescent period. Nevertheless, race- and sex-specific growth curves of pulmonary function versus height make it easy to display and evaluate repeated measures of pulmonary function for an individual child 29.

Details of reference populations and regression equations for children and adolescents are summarised by Quanjer et al. 30. Lung volume reference equations have been frequently derived from relatively small populations (<200 children) over a 6–12-yr age range when growth and developmental changes are extremely rapid. Relatively few studies have taken puberty or age into account.

A comprehensive listing of published reference equations for lung volumes was published in 1983 by the ECCS 20 and updated in 1993 8. A set of equations was created by combining the equations in this list with the intent to use the combined equations for adults aged 18–70 yrs with a height range of 155–195 cm in males, and 145–180 cm in females.

A report on an ATS workshop on lung volume measurements 7 reviewed published reference values in infants, pre-school children, children, adolescents and adults, and gave recommendations for selecting reference values, expressing results, measuring ancillary variables and designing future studies. Most reference equations for children are derived from Caucasian populations.

Differences due to ethnicity are not well defined 3336. These differences may be explained, in part, by differences in trunk length relative to standing height, but there are also differences in fat-free mass, chest dimensions and strength of respiratory muscles. Until better information is available, correction factors for Black and Asian children could be the same as those recommended for adults 7. Reference values for RV, VC and TLC are, on average, 12% lower in Blacks than in Whites 35; this difference may be smaller in elderly persons than in young adults 36. Reference values for absolute lung volumes for adults of Asian ethnicity are generally considered to be lower than for Whites, but the magnitude of the differences is not well defined, and the difference may be less in Asians raised on “Western” diets during childhood 37. According to the ATS 1991 document 5, no race correction is used for TLC or RV in Hispanic or Native American subjects in North America. For African American, Asian American and East Indian subjects, race corrections of 0.88 for TLC, FRC, and 0.93 for RV, are used. Race corrections should not be used for RV/TLC.

Table 2 reports studies on reference equations published from 1993 to August 2004, and equations derived from a MEDLINE search under the keywords “reference equations” and “lung volumes”. Its purpose is to recognise and encourage the continuing interest of worldwide researchers in deriving and using reference equations.

In practice, many USA and European laboratories use the reference equations for TLC, FRC and RV recommended by the 1995 ATS/ERS workshop 7 or by the ECCS 8, 20.

Diffusing capacity for carbon monoxide

Selecting reference values for DL,CO is more problematic than selecting reference values for spirometry because inter-laboratory differences are much larger for DL,CO 38, 39. Some of these differences can be attributed to the method of calculating DL,CO and adjusting for haemoglobin concentration, carboxyhaemoglobin concentration and altitude. Laboratory directors should thoughtfully select reference values that match the numbers produced in their laboratories. Optimally, it would require individual laboratories to measure DL,CO in a sample of healthy subjects and compare the results with several reference equations. At the very least, laboratory directors should be alert to frequent interpretations that do not match the clinical situation. Such mismatches may signal inappropriate reference values or problems with the DL,CO measurement.

Predicted values for alveolar volume (VA) inspired volume (VI), DL,CO and transfer coefficient of the lung for carbon monoxide (KCO) should be derived from the same source. As DL,CO and KCO may be variably affected by factors previously described in this series of Task Force reports 4, a statement should be included describing which parameters might have been used to adjust the predicted values (e.g. VA, haemoglobin and carboxyhaemoglobin concentrations, and altitude).

Table 3 shows studies on reference equations, published from 1995 to August 2004, and some derived from a MEDLINE search under the keywords “reference equations” and “diffusing capacity” or “diffusion”. Its purpose is to recognise and encourage the continuing interest of worldwide researchers in deriving and using reference equations.

A single “summary” prediction equation was proposed by the ERS 38 and suggested by the ATS 39. At present, however, a single equation set for DL,CO cannot be recommended because of the relatively high inter-laboratory variability. Commonly used equations appear to be those from the 1993 ERS document 38 and those of Crapo and Morris 40. In Europe, equations from Cotes et al. 41, Paoletti et al. 42, and Roca et al. 43 are also used.

Table 4 provides a summary of the reference values used for general issues, spirometry, lung volumes and diffusing capacity.


General issues

PFT interpretations should be clear, concise and informative. A mere statement of which values are normal or low is not helpful. Ideally, the principles of clinical decision making should be applied to the interpretation of the results of PFTs 44, where the post-test probability of disease is estimated after taking into consideration the pre-test probability of disease, the quality of the test results, the downside of a false-positive and false-negative interpretation, and, finally, the test results themselves and how they compare with reference values. This is often not possible because many, if not most, tests are interpreted in the absence of any clinical information. To improve this situation, it may be useful, whenever possible, to ask the physicians who are responsible for ordering tests to state the clinical question to be answered and, before testing, ask patients why they were sent for testing. Similarly, recording respiratory symptoms, such as cough, phlegm, wheezing and dyspnoea, as well as smoking status, and recent bronchodilator use could be helpful in this regard.

The interpretation will be most meaningful if the interpreter can address relevant clinical diagnoses, the chest radiograph appearance, the most recent haemoglobin value, and any suspicion of neuromuscular disease or upper airway obstruction (UAO).

Obstructive abnormalities

An obstructive ventilatory defect is a disproportionate reduction of maximal airflow from the lung in relation to the maximal volume (i.e. VC) that can be displaced from the lung 4547. It implies airway narrowing during exhalation and is defined by a reduced FEV1/VC ratio below the 5th percentile of the predicted value. A typical example is shown in figure 1a.

The earliest change associated with airflow obstruction in small airways is thought to be a slowing in the terminal portion of the spirogram, even when the initial part of the spirogram is barely affected 4547. This slowing of expiratory flow is most obviously reflected in a concave shape on the flow–volume curve. Quantitatively, it is reflected in a proportionally greater reduction in the instantaneous flow measured after 75% of the FVC has been exhaled (FEF75%) or in mean expiratory flow between 25% and 75% of FVC than in FEV1. However, abnormalities in these mid-range flow measurements during a forced exhalation are not specific for small airway disease in individual patients 48.

As airway disease becomes more advanced and/or more central airways become involved, timed segments of the spirogram such as the FEV1 will, in general, be reduced out of proportion to the reduction in VC.

Special attention must be paid when FEV1 and FVC are concomitantly decreased and the FEV1/FVC ratio is normal or almost normal. This pattern most frequently reflects failure of the patient to inhale or exhale completely. It may also occur when the flow is so slow that the subject cannot exhale long enough to empty the lungs to RV. In this circumstance, the flow–volume curve should appear concave toward the end of the manoeuvre. TLC will be normal and FEF75 will be low. Measurement of slow VC (inspiratory or expiratory) may then give a more correct estimate of the FEV1/VC ratio. Another possible cause of this pattern is patchy collapse of small airways early in exhalation 8, 4952. Under these conditions, TLC may be normal, but RV is ordinarily increased. A typical example is shown in figure 1b. When this pattern is observed in a patient performing a maximal, sustained effort, it may be useful to repeat spirometry after treatment with an inhaled bronchodilator. Significant improvement in the FEV1, FVC or both would suggest the presence of reversible airflow obstruction.

Apart from this unusual circumstance, measurement of lung volumes is not mandatory to identify an obstructive defect. It may, however, help to disclose underlying disease and its functional consequences. For example, an increase in TLC, RV or the RV/TLC ratio above the upper limits of natural variability may suggest the presence of emphysema, bronchial asthma or other obstructive diseases 47, as well as the degree of lung hyperinflation.

Airflow resistance is rarely used to identify airflow obstruction in clinical practice. It is more sensitive for detecting narrowing of extrathoracic or large central intrathoracic airways than of more peripheral intrathoracic airways 47. It may be useful in patients who are unable to perform a maximal forced expiratory manoeuvre.

Restrictive abnormalities

A restrictive ventilatory defect is characterised by a reduction in TLC below the 5th percentile of the predicted value, and a normal FEV1/VC. A typical example is shown in figure 1c. The presence of a restrictive ventilatory defect may be suspected when VC is reduced, the FEV1/VC is increased (>85–90%) and the flow–volume curve shows a convex pattern. Once again, the pattern of a reduced VC and a normal or even slightly increased FEV1/VC is often caused by submaximal inspiratory or expiratory efforts and/or patchy peripheral airflow obstruction, and a reduced VC by itself does not prove a restrictive ventilatory defect. It is associated with a low TLC no more than half the time 53, 54.

Pneumothorax and noncommunicating bullae are special cases characterised by a normal FEV1/VC and TLC measured in a body plethysmograph, but low FEV1 and VC values. In these conditions, TLC assessed by gas dilution techniques will be low.

A low TLC from a single-breath test (such as VA from the DL,CO test) should not be interpreted as demonstrating restriction, since such measurements systematically underestimate TLC 55. The degree of underestimation increases as airflow obstruction worsens. In the presence of severe airflow obstruction, TLC can be underestimated by as much as 3 L, greatly increasing the risk of misclassification of the type of PFT abnormality 55, 56. A method of adjusting the single-breath VA for the effect of airway obstruction has been published, but needs further validation 57.

Mixed abnormalities

A mixed ventilatory defect is characterised by the coexistence of obstruction and restriction, and is defined physiologically when both FEV1/VC and TLC are below the 5th percentiles of their relevant predicted values. Since VC may be equally reduced in both obstruction and restriction, the presence of a restrictive component in an obstructed patient cannot be inferred from simple measurements of FEV1 and VC. A typical example is presented in figure 1d. If FEV1/VC is low and the largest measured VC (pre- or post-bronchodilator VC or VI in the DL,CO test) is below its lower limits of normal (LLN), and there is no measurement of TLC by body plethysmography, one can state that the VC was also reduced, probably due to hyperinflation, but that a superimposed restriction of lung volumes cannot be ruled out 58. Conversely, when FEV1/VC is low and VC is normal, a superimposed restriction of lung volumes can be ruled out 53, 54.

Table 5 shows a summary of the types of ventilatory defects and their diagnoses.


The definition of an obstructive pulmonary defect given in the present document is consistent with the 1991 ATS statement on interpretation 5, but contrasts with the definitions suggested by both Global Initiative for Chronic Obstructive Lung Disease (GOLD) 59 and ATS/ERS guidelines on COPD 60, in that FEV1 is referred to VC rather than just FVC and the cut-off value of this ratio is set at the 5th percentile of the normal distribution rather than at a fixed value of 0.7. This committee feels that the advantage of using VC in place of FVC is that the ratio of FEV1 to VC is capable of accurately identifying more obstructive patterns than its ratio to FVC, because FVC is more dependent on flow and volume histories 61. In contrast with a fixed value of 0.7, the use of the 5th percentile does not lead to an overestimation of the ventilatory defect in older people with no history of exposure to noxious particles or gases 62.

The assumption that a decrease in major spirometric parameters, such as FEV1, VC, FEV1/VC and TLC, below their relevant 5th percentiles is consistent with a pulmonary defect is a useful simple approach in clinical practice. Problems arise, however, when some or all of these variables lie near their upper limits of normal or LLN. In these cases, a literal interpretation of the functional pattern is too simplistic and could fail to properly describe the functional status.

The current authors suggest that additional studies should be done in these circumstances if they are indicated by the clinical problem being addressed. Such tests could include bronchodilator response, DL,CO, gas-exchange evaluation, measurement of respiratory muscle strength or exercise testing.

Caution is also recommended when TLC is at the LLN and coexists with a disease expected to lead to lung restriction. A typical example is lung resection. The expected restrictive defect would be difficult to prove on the simple basis of TLC as per cent of predicted if the latter remains above the 5th percentile of predicted as a result of subsequent lung growth or of a large TLC before surgery. Similar care must be taken in cases where diseases with opposing effects on TLC coexist, such as interstitial lung disease (ILD) and emphysema.

While patterns of physiological abnormalities can be recognised, they are seldom pathognomonic for a specific disease entity. The types of clinical illness most likely to produce an observed set of physiological disturbances can be pointed out. Regardless of the extent of testing, it is important to be conservative in suggesting a specific diagnosis for an underlying disease process based only on pulmonary function abnormalities.

The VC, FEV1, FEV1/VC ratio and TLC are the basic parameters used to properly interpret lung function (fig. 2). Although FVC is often used in place of VC, it is preferable to use the largest available VC, whether obtained on inspiration (IVC), slow expiration (SVC) or forced expiration (i.e. FVC). The FVC is usually reduced more than IVC or SVC in airflow obstruction 61. The FEV6 may be substituted for VC if the appropriate LLN for the FEV1/FEV6 is used (from the NHANES III equations) 12, 63. Limiting primary interpretation of spirograms to VC, FEV1 and FEV1/VC avoids the problem of simultaneously examining a multitude of measurements to see if any abnormalities are present, a procedure leading to an inordinate number of “abnormal” tests, even among the healthiest groups in a population 64, 65. When the rate of abnormality for any single test is only 5%, the frequency of at least one abnormal test was shown to be 10% in 251 healthy subjects when the FEV1, FVC and FEV1/FVC ratio were examined and increased to 24% when a battery of 14 different spirometric measurements were analysed 23. It should be noted, however, that additional parameters, such as the peak expiratory flow (PEF) and maximum inspiratory flows, may assist in diagnosing extrathoracic airway obstruction.

The most important parameter for identifying an obstructive impairment in patients is the FEV1/VC ratio. In patients with respiratory diseases, a low FEV1/VC, even when FEV1 is within the normal range, predicts morbidity and mortality 66. For healthy subjects, the meaning of a low FEV1/FVC ratio accompanied by an FEV1 within the normal range is unclear. This pattern is probably due to “dysanaptic” or unequal growth of the airways and lung parenchyma 67 (referred to in a previous ATS document as a possible physiological variant when FEV1 was ≥100% pred 5). Whether this pattern represents airflow obstruction will depend on the prior probability of obstructive disease and possibly on the results of additional tests, such as bronchodilator response, DL,CO, gas-exchange evaluation, and measurement of muscle strength or exercise testing. Expiratory flow measurements other than the FEV1 and FEV1/VC should be considered only after determining the presence and clinical severity of obstructive impairment using the basic values mentioned previously. When the FEV1 and FEV1/VC are within the expected range, the clinical significance of abnormalities in flow occurring late in the maximal expiratory flow–volume curve is limited. In the presence of a borderline value of FEV1/VC, however, these tests may suggest the presence of airway obstruction. The same is true for average flows, such as mid-expiratory flow (MEF25–75%), especially in children with cystic fibrosis 68, 69. Even with this limited use, the wide variability of these tests in healthy subjects must be taken into account in their interpretation.

The maximal voluntary ventilation (MVV) is not generally included in the set of lung function parameters necessary for diagnosis or follow-up of the pulmonary abnormalities because of its good correlation with FEV1 70. However, it may be of some help in clinical practice. For example, a disproportionate decrease in MVV relative to FEV1 has been reported in neuromuscular disorders 71, 72 and UAO 73. In addition, it is also used in estimating breathing reserve during maximal exercise 74, although its application may be of limited value in mild-to-moderate COPD 75, 76. For these purposes, the current authors suggest that MVV should be measured rather than estimated by multiplying FEV1 by a constant value, as is often done in practice.


A method of categorising the severity of lung function impairment based on the FEV1 % pred is given in table 6. It is similar to several previous documents, including GOLD 59, ATS 1986 77, ATS 1991 5, and the American Medical Association (AMA) 78. The number of categories and the exact cut-off points are arbitrary.

Severity scores are most appropriately derived from studies that relate pulmonary function test values to independent indices of performance, such as ability to work and function in daily life, morbidity and prognosis 7982. In general, the ability to work and function in daily life is related to pulmonary function, and pulmonary function is used to rate impairment in several published systems 7779, 83. Pulmonary function level is also associated with morbidity, and the patients with lower function have more respiratory complaints 82.

Lung function level is also associated with prognosis, including a fatal outcome from heart as well as lung disease 84, 85, even in patients who have never smoked 86. In the Framingham study, VC was a major independent predictor of cardiovascular morbidity and mortality 84, 85. In several occupational cohorts, FEV1 and FEV1/FVC were independent predictors of all-cause or respiratory disease mortality 8789. In addition, a meta-analysis of mortality in six surveys in various UK working populations showed that the risk of dying from COPD was related to the FEV1 level. In comparison to those whose FEV1 at an initial examination was within 1 sd of average, those whose FEV1 was >2 sd below average were 12 times more likely to die of COPD, more than 10 times as likely to die of non-neoplastic respiratory disease, and more than twice as likely to die of vascular disease over a 20-yr follow-up period 90. Although there is good evidence that FEV1 correlates with the severity of symptoms and prognosis in many circumstances 79, 82, 90, the correlations do not allow one to accurately predict symptoms or prognosis for individual patients.

The DL,CO is also an important predictor of mortality both in the general population 91 and in patients after pulmonary resection 92.

Though the FEV1 % pred is generally used to grade severity in patients with obstructive, restrictive and mixed pulmonary defects, it has little applicability to patients with UAO, such as tracheal stenosis, where obstruction could be life-threatening and yet be classified as mildly reduced by this scheme. In addition, there is little data documenting the performance of other functional indexes, such as FRC in airflow obstruction or TLC in lung restriction as indices to categorise severity of impairment.

VC is reduced in relation to the extent of loss of functioning lung parenchyma in many nonobstructive lung disorders. It is also of some use in assessing respiratory muscle involvement in certain neuromuscular diseases. VC may be only slightly impaired in diffuse interstitial diseases of sufficient severity to lead to marked loss of diffusing capacity and severe blood gas abnormalities 63. The onset of a severe respiratory problem in patients with a rapidly progressive neuromuscular disease may be associated with only a small decrement in VC 47, 93.

FEV1 and FVC may sometimes fail to properly identify the severity of ventilatory defects, especially at the very severe stage for multiple reasons. Among them are the volume history effects of the deep breath preceding the forced expiratory manoeuvre on the bronchial tone and, thus, calibre 9498, and the inability of these parameters to detect whether tidal breathing is flow limited or not 99102. The FEV1/VC ratio should not be used to determine the severity of an obstructive disorder, until new research data are available. Both the FEV1 and VC may decline with the progression of disease, and an FEV1/VC of 0.5/1.0 indicates more impairment than one of 2.0/4.0, although the ratio of both is 50%. While the FEV1/VC ratio should not routinely be used to determine the severity of an obstructive disorder, it may be of value when persons having genetically large lungs develop obstructive disease. In these cases, the FEV1/VC ratio may be very low (60%), when the FEV1 alone is within the mild category of obstruction (i.e. >70% pred).

Recent studies have stressed the importance of additional measurements in assessing the severity of the disease. For example, when airflow obstruction becomes severe, FRC, RV, TLC and RV/TLC tend to increase as a result of decreased lung elastic recoil and/or dynamic mechanisms 47, 103, 104. The degree of hyperinflation parallels the severity of airway obstruction 58. On one hand, lung hyperinflation is of benefit because it modulates airflow obstruction, but, on the other hand, it causes dyspnoea because of the increased elastic load on inspiratory muscles 47. In a recent investigation, resting lung hyperinflation, measured as inspiratory capacity (IC)/TLC, was an independent predictor of respiratory and all-cause mortality in COPD patients 105. In addition, in either severe obstructive or restrictive diseases, tidal expiratory flow often impinges on maximum flow 98, 99, 102. This condition, denoted as expiratory flow limitation during tidal breathing (EFL), is relatively easy to measure in practice by comparing tidal and forced expiratory flow–volume loops. Its clinical importance is that it contributes to increased dyspnoea 100, puts the inspiratory muscles at a mechanical disadvantage 43 and causes cardiovascular side-effects 106. Although there currently isn't sufficient evidence to recommend the routine use of measurements of hyperinflation or EFL to score the severity of lung function impairment, they may be helpful in patients with disproportionate differences between spirometric impairment and dyspnoea.

Finally, the reported increase in RV in obstruction is deemed to be a marker of airway closure 47, 103. Although its clinical relevance remains uncertain, especially with regard to assessment of severity, RV may be useful in special conditions, including predicting the likelihood of lung function improvement after lung volume-reduction surgery 104.

Table 7 shows the summary of the considerations for severity classification.


Bronchial responsiveness to bronchodilator medications is an integrated physiological response involving airway epithelium, nerves, mediators and bronchial smooth muscle. Since the within-individual difference in response to a bronchodilator is variable, the assumption that a single test of bronchodilator response is adequate to assess both the underlying airway responsiveness and the potential for therapeutic benefits of bronchodilator therapy is overly simplistic 107. Therefore, the current authors feel that the response to a bronchodilator agent can be tested either after a single dose of a bronchodilator agent in the PFT laboratory or after a clinical trial conducted over 2–8 weeks.

The correlation between bronchoconstriction and bronchodilator response is imperfect, and it is not possible to infer with certainty the presence of one from the other.

There is no consensus about the drug, dose or mode of administering a bronchodilator in the laboratory. However, when a metered dose inhaler is used, the following procedures are suggested in order to minimise differences within and between laboratories. Short-acting β2-agonists, such as salbutamol, are recommended. Four separate doses of 100 µg should be used when given by metered dose inhaler using a spacer. Tests should be repeated after a 15-min delay. If a bronchodilator test is performed to assess the potential therapeutic benefits of a specific drug, it should be administered in the same dose and by the same route as used in clinical practice, and the delay between administration and repeated spirometric measurements should reflect the reported time of onset for that drug.

The first step in interpreting any bronchodilator test is to determine if any change greater than random variation has occurred. The per cent change in FVC and FEV1 after bronchodilator administration in general population studies 108110 and patient populations 101, 111113 are summarised in table 8. Studies show a tendency for the calculated bronchodilator response to increase with decreasing baseline VC or FEV1, regardless of whether the response was considered as an absolute change or as a per cent of the initial value. Bronchodilator responses in patient-based studies are, therefore somewhat higher than those in general population studies.

There is no clear consensus about what constitutes reversibility in subjects with airflow obstruction 111, 114. In part, this is because there is no consensus on how a bronchodilator response should be expressed, the variables to be used, and, finally, the kind, dose and inhalation mode of bronchodilator agent. The three most common methods of expressing bronchodilator response are per cent of the initial spirometric value, per cent of the predicted value, and absolute change.

Expressing the change in FEV1 and/or FVC as a per cent of predicted values has been reported to have advantages over per cent change from baseline 115. When using per cent change from baseline as the criterion, most authorities require a 12–15% increase in FEV1 and/or FVC as necessary to define a meaningful response. Increments of <8% (or <150 mL) are likely to be within measurement variability 107, 115. The current authors recommend using the per cent change from baseline and absolute changes in FEV1 and/or FVC in an individual subject to identify a positive bronchodilator response. Values >12% and 200 mL compared with baseline during a single testing session suggest a “significant” bronchodilatation. If the change in FEV1 is not significant, a decrease in lung hyperinflation may indicate a significant response 101. The lack of a response to bronchodilator testing in a laboratory does not preclude a clinical response to bronchodilator therapy.

The MEF25–75% is a highly variable spirometric test, in part because it depends on FVC, which increases with expiratory time in obstructed subjects. If FVC changes, post-bronchodilator MEF25–75% is not comparable with that measured before the bronchodilator. Volume adjustment of MEF25–75% has been proposed to solve this problem 116, 117. At least two studies have assessed the utility of MEF25–75%. The results were disappointing; only 8% of asthmatics 117 and 7% of patients with COPD were identified as outside the expected range by MEF25–75% criteria alone. Tests such as the FEV1/VC ratio and instantaneous flows measured at some fraction of the VC may also be misleading in assessing bronchodilator response if expiratory time changes are not considered and if flows are not measured at the same volume below TLC.

If the change is above the threshold of natural variability, then the next step is to determine if this change is clinically important. This aspect of the interpretation is harder to define and depends on the reasons for undertaking the test. For instance, even if asthmatics tend to show a larger increase in flow and volume after inhaling a dilator agent than COPD patients, the response to a bronchodilator has never been shown to be capable of clearly separating the two classes of patients 101, 109, 111, 114. In addition, it must be also acknowledged that responses well below the significant thresholds may be associated with symptom improvement and patient performance 118. The possible reasons are discussed as follows.

Quite often, responses to bronchodilator therapy are unpredictably underestimated by FEV1 and/or FVC in comparison to airway resistance or flow measured during forced expiratory manoeuvres initiated from a volume below TLC (partial expiratory flow–volume manoeuvres) in both healthy subjects and patients with chronic airflow obstruction 8, 101, 102, 119122. These findings are probably due to the fact that deep inhalations tend to reduce airway calibre, especially after a bronchodilator 101, 120. In patients with airflow obstruction, the increase in expiratory flow after bronchodilation is often associated with a decrease in FRC or an increase in IC of similar extent at rest and during exercise 101, 123. The improvement of the lung function parameters in the tidal breathing range and not following a deep breath may explain the decrease in shortness of breath after inhaling a bronchodilator, despite no or minimal changes in FEV1 and/or FVC. Short-term intra-individual variabilities for partial flows and IC have been reported 101. Therefore, the lack of increase of FEV1 and/or FVC after a bronchodilator is not a good reason to avoid 1–8-week clinical trial with bronchoactive medication.

An isolated increase in FVC (>12% of control and >200 mL) not due to increased expiratory time after salbutamol is a sign of bronchodilation 124. This may, in part, be related to the fact that deep inhalations tend to reduce airway calibre and/or airway wall stiffness, especially after a bronchodilator 101, 120.

Table 9 shows a summary of the suggested procedures for laboratories relating to bronchodilator response.


Central airway obstruction and UAO may occur in the extrathoracic (pharynx, larynx, and extrathoracic portion of the trachea) and intrathoracic airways (intrathoracic trachea and main bronchi). This condition does not usually lead to a decrease in FEV1 and/or VC, but PEF can be severely affected. Therefore, an increased ratio of FEV1 divided by PEF (mL·L−1·min−1) can alert the clinician to the need for an inspiratory and expiratory flow–volume loop 125. A value >8 suggests central or upper airway obstruction may be present 126. Poor initial effort can also affect this ratio.

At least three maximal and repeatable forced inspiratory and forced expiratory flow–volume curves are necessary to evaluate for central or upper airway obstruction. It is critical that the patient's inspiratory and expiratory efforts are near maximal and the technician should confirm this in the quality notes. When patient effort is good, the pattern of a repeatable plateau of forced inspiratory flow, with or without a forced expiratory plateau, suggests a variable extrathoracic central or upper airway obstruction (fig. 3). Conversely, the pattern of a repeatable plateau of forced expiratory flow, along with the lack of a forced inspiratory plateau suggests a variable, intrathoracic central or upper airway obstruction. The pattern of a repeatable plateau at a similar flow in both forced inspiratory and expiratory flows suggests a fixed central or upper airway obstruction (fig. 3).

In general, maximum inspiratory flow is largely decreased with an extrathoracic airway obstruction, because the pressure surrounding the airways (which is almost equal to atmospheric) cannot oppose the negative intraluminal pressure generated with the inspiratory effort. In contrast, it is little affected by an intrathoracic airway obstruction, for the pressure surrounding the intrathoracic airways (which is close to pleural pressure) strongly opposes the negative intraluminal pressure on inspiration, thus limiting the effects of the obstruction on flow. With unilateral main bronchus obstruction, a rare event, maximum inspiratory flow tends to be higher at the beginning than towards the end of the forced inspiration because of a delay in gas filling (fig. 4).

Maximum expiratory flow at high lung volume (especially peak flow) is generally decreased in both intrathoracic and extrathoracic lesions 126129. In contrast, maximum flows may be normal in the presence of a variable lesion, such as vocal cord paralysis. Flow oscillations (saw-tooth pattern) may be occasionally observed on the either inspiratory or expiratory phase, and probably represent a mechanical instability of the airway wall.

The effects of anatomical or functional lesions on maximum flows depend on the site of the obstruction, kind of lesion (variable or fixed) and the extent of anatomical obstruction 61, 127, 130. Typical cases of extra- and intrathoracic central or upper airway obstruction are reported in figures 3 and 4. The absence of classic spirometric patterns for central airway obstruction does not accurately predict the absence of pathology. As a result, clinicians need to maintain a high degree of suspicion for this problem, and refer suspected cases for visual inspection of the airways. The authors feel that, although maximum inspiratory and expiratory flow–volume loops are of great help to alert clinicians to the possibility of central or upper airway obstruction, endoscopic and radiological techniques are the next step to confirm the dysfunction.

The parameters presented in table 10 may help to distinguish intrathoracic from extrathoracic airway obstructions.

Table 11 gives a summary of the relevant issues concerning UAO.


Evaluation of an individual's change in lung function following an intervention or over time is often more clinically valuable than a single comparison with external reference (predicted) values. It is not easy to determine whether a measured change reflects a true change in pulmonary status or is only a result of test variability. All lung function measurements tend to be more variable when made weeks to months apart than when repeated at the same test session or even daily 25, 131. The short-term repeatability of tracked parameters should be measured using biological controls. This is especially important for the DL,CO 132, 133, since small errors in measurements of inspiratory flows or exhaled gas concentrations translate into large DL,CO errors. The variability of lung volume measurements has recently been reviewed 134.

The optimal method of expressing the short-term variability (measurement noise) is to calculate the coefficient of repeatability (CR) instead of the more popular coefficient of variation 135. Change measured for an individual patient that falls outside the CR for a given parameter may be considered significant. The CR may be expressed as an absolute value (such as 0.33 L for FEV1 or 5 units for DL,CO) 136 or as a percentage of the mean value (such as 11% for FEV1) 137.

It is more likely that a real change has occurred when more than two measurements are performed over time. As shown in table 12, significant changes, whether statistical or biological, vary by parameter, time period and the type of patient. When there are only two tests available to evaluate change, the large variability necessitates relatively large changes to be confident that a significant change has in fact occurred. Thus, in subjects with relatively “normal” lung function, year-to-year changes in FEV1 over 1 yr should exceed 15% before confidence can be given to the opinion that a clinically meaningful change has occurred 5.

For tracking change, FEV1 has the advantage of being the most repeatable lung function parameter and one that measures changes in both obstructive and restrictive types of lung disease. Two-point, short-term changes of >12% and >0.2 L in the FEV1 are usually statistically significant and may be clinically important. Changes slightly less than these may, perhaps, be equally significant, depending on the reproducibility of the pre- and post-bronchodilator results. Other parameters such as VC, IC, TLC and DL,CO may also be tracked in patients with ILD or severe COPD 138, 140142. Tests like VC and FVC may be relevant to COPD because they may increase when FEV1 does not, and changes in DL,CO, in the absence of change in spirometry variables, may be clinically important. Again, when too many indices of lung function are tracked simultaneously, the risk of false-positive indications of change increases.

The clinician seeing the patient can often interpret results of serial tests in a useful manner, which is not reproducible by any simple algorithm. Depending on the clinical situation, statistically insignificant trends in lung function may be meaningful to the clinician. For example, seemingly stable test results may provide reassurance in a patient receiving therapy for a disease that is otherwise rapidly progressive. The same test may be very disappointing if one is treating a disorder that is expected to improve dramatically with the therapy prescribed. Conversely, a statistically significant change may be of no clinical importance to the patient. The largest errors occur in attempting to interpret serial changes in subjects without disease, because test variability will usually far exceed the true annual decline, and reliable rates of change for an individual subject cannot be calculated without prolonged follow-up 143.

Test variability can be reduced when lung function standards and guidelines are followed strictly. Simple plots (i.e. trending) of lung function with time can provide additional information to help differentiate true change in lung function from noise. Measuring decline in lung function as a means of identifying individuals (such as smokers) who are losing function at excessive rates has been proposed. However, establishing an accelerated rate of loss in an individual is very difficult, and requires many measurements over several years with meticulous quality control of the measurements.

Table 13 shows a summary of the considerations involved in interpreting lung function changes.


The lower 5th percentile of the reference population should be used as LLN for DL,CO and KCO (if the latter is used). Table 14 presents a scheme to grade the severity of reductions in DL,CO.

The pathophysiological importance of this test has been recently reviewed 144, 145.

Interpreting the DL,CO, in conjunction with spirometry and lung volumes assessment, may assist in diagnosing the underlying disease (fig. 2). For instance, normal spirometry and lung volumes associated with decreased DL,CO may suggest anaemia, pulmonary vascular disorders, early ILD or early emphysema. In the presence of restriction, a normal DL,CO may be consistent with chest wall or neuromuscular disorders, whereas a decrease suggests ILDs. In the presence of airflow obstruction, a decreased DL,CO suggests emphysema 146, but airway obstruction and a low DL,CO are also seen in lymphangioleiomyomatosis 147. Patients with ILD, sarcoidosis and pulmonary fibrosis usually have a low DL,CO 135137, 140. A low DL,CO is also seen in patients with chronic pulmonary embolism, primary pulmonary hypertension 148, and other pulmonary vascular diseases. These patients may or may not also have restriction of lung volumes 149.

A high DL,CO is associated with asthma 150, obesity 151 and intrapulmonary haemorrhage 152.

Adjustments of DL,CO for changes in haemoglobin and carboxyhaemoglobin are important, especially in situations where patients are being monitored for possible drug toxicity, and where haemoglobin is subject to large shifts (e.g. chemotherapy for cancer).

Adjusting DL,CO for lung volume using DL,CO/VA or DL,CO/TLC is controversial 153, 154. Conceptually, a loss of DL,CO that is much less than a loss of volume (low DL,CO but high DL,CO/VA) might suggest an extraparenchymal abnormality, such as a pneumonectomy or chest wall restriction, whereas a loss of DL,CO that is much greater than a loss of volume (low DL,CO and low DL,CO/VA) might suggest parenchymal abnormalities. The relationship between DL,CO and lung volume, however, is not linear and markedly less than 1:1, so these simple ratios as traditionally reported do not provide an appropriate way to normalise DL,CO for lung volume 154159. Nonlinear adjustments may be considered, but their clinical utility must be established before they can be recommended. Meanwhile, it is advisable to keep examining DL,CO/VA and VA separately 153, in so far as it may provide information on disease pathophysiology that cannot be obtained from their product, the DL,CO.

Table 15 shows a summary on the considerations for DL,CO interpretation.


Table 16 contains a list of abbreviations and their meanings, which have been used in this series of Task Force reports.

Fig. 1—

a, b) Examples of obstructive pulmonary defects with a low (a; forced expiratory volume in one second (FEV1) 38%; FEV1/vital capacity (VC) 46%; peak expiratory flow (PEF) 48%; total lung capacity (TLC) 101%) or normal (b; FEV1 57%; FEV1/VC 73%; PEF 43%; TLC 96%) ratio of FEV1/VC. In both cases, TLC is normal, and flows are less than expected over the entire volume range. c) Example of a typical restrictive defect (FEV1 66%; FEV1/VC 80%; PEF 79%; TLC 62%). The TLC is low and flow is higher than expected at a given lung volume. d) Example of a typical mixed defect characterised by a low TLC and a low FEV1/VC ratio (FEV1 64%; FEV1/VC 64%; PEF 82%; TLC 72%). – – – –: predicted flow–volume curves; ––––: observed inspiratory and expiratory flow–volume curves (as indicated in a).

Fig. 2—

A simplified algorithm that may be used to assess lung function in clinical practice. It presents classic patterns for various pulmonary disorders. As in any such diagram, patients may or may not present with the classic patterns, depending on their illnesses, severity and lung function prior to the disease onset (e.g. did they start with a vital capacity (VC) close to the upper or lower limits of normal (LLN)). The decisions about how far to follow this diagram are clinical, and will vary depending on the questions being asked and the clinical information available at the time of testing. The forced expiratory volume in one second (FEV1)/VC ratio and VC should be considered first. Total lung capacity (TLC) is necessary to confirm or exclude the presence of a restrictive defect when VC is below the LLN. The algorithm also includes diffusing capacity for carbon monoxide (DL,CO) measurement with the predicted value adjusted for haemoglobin. In the mixed defect group, the DL,CO patterns are the same as those for restriction and obstruction. This flow chart is not suitable for assessing the severity of upper airway obstruction. PV: pulmonary vascular; CW: chest wall; NM: neuromuscular; ILD: interstitial lung diseases; CB: chronic bronchitis.

Fig. 3—

Idealised examples of a) fixed, b) variable extrathoracic, and c) variable intrathoracic airway obstruction.

Fig. 4—

Example of unilateral main bronchus obstruction due to a valve-like mechanism occluding the main left stem bronchus during inspiration as a result of a surgical scar. There is a delay in gas filling towards the end of the forced inspiration as evidence of the variable unilateral main bronchus obstruction (forced expiratory volume in one second (FEV1): 76%; FEV1/vital capacity: 70%; peak expiratory flow: 93%; total lung capacity: 80%). -----: predicted expiratory flow–volume loop; ––––: recorded maximum inspiratory and expiratory flow–volume loops.

View this table:
Table. 1—

Outcomes of a MEDLINE search using the keywords“reference equations” and “spirometry”

View this table:
Table. 2—

Outcomes of a MEDLINE search under the keywords“reference equations” and “lung volumes”

View this table:
Table. 3—

Outcome of a MEDLINE search under the keywords“reference equations” and “diffusing capacity” or “diffusion”

View this table:
Table. 4—

Summary of the usage of reference values

View this table:
Table. 5—

Types of ventilatory defects and their diagnoses

View this table:
Table. 6—

Severity of any spirometric abnormality based on the forced expiratory volume in one second(FEV1)

View this table:
Table. 7—

Summary of the considerations for severity classification

View this table:
Table. 8—

Selected studies of bronchodilator response

View this table:
Table. 9—

Summary of the procedures relating to bronchodilator response

View this table:
Table. 10—

Lung function parameters capable of differentiating extrathoracic from intrathoracic obstruction

View this table:
Table. 11—

Summary of the issues concerning central or upper airway obstruction

View this table:
Table. 12—

Reported significant changes in forced vital capacity(FVC), forced expiratory volume in one second (FEV1), mid-expiratory flow (MEF25–75%) and carbon monoxide diffusing capacity (DL,CO) over time

View this table:
Table. 13—

Summary of the considerations for the interpretation of change in lung function

View this table:
Table. 14—

Degree of severity of decrease in diffusing capacity for carbon monoxide(DL,CO)

View this table:
Table. 15—

Summary of the considerations for diffusing capacity for carbon monoxide(DL,CO) interpretation

View this table:
Table. 16—

List of abbreviations and meanings


R. Pellegrino: Azienda Ospedaliera S. Croce e Carle, Cuneo, Italy; G. Viegi: CNR Institute of Clinical Physiology, Pisa, Italy; V. Brusasco: Università degli Studi di Genova, Genova, Italy; R.O. Crapo and R. Jensen: LDS Hospital, Salt Lake City, UT, USA; F. Burgos: Hospital Clinic Villarroel, Barcelona, Spain; R. Casaburi: Harbor UCLA Medical Center, Torrance, CA, USA; A. Coates: Hospital for Sick Children, Toronto, ON, Canada; C.P.M. van der Grinten: University Hospital of Maastrict, Maastrict, the Netherlands; P. Gustafsson: Queen Silvias Children's Hospital, Gothenburg, Sweden; J. Hankinson: Hankinson Consulting, Inc., Valdosta, GA, USA; D.C. Johnson: Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA; N. MacIntyre: Duke University Medical Center, Durham, NC, USA; R. McKay: Occupational Medicine, Cincinnati, OH, USA; M.R. Miller: University Hospital Birmingham NHS Trust, Birmingham, UK; D. Navajas: Lab Biofisica I Bioenginyeria, Barcelona, Spain; O.F. Pedersen: University of Aarhus, Aarhus, Denmark; J. Wanger: Pharmaceutical Research Associates, Inc., Lenexa, KS, USA.


  • Previous articles in this series: No. 1: Miller MR, Crapo R, Hankinson J, et al. General considerations for lung function testing. Eur Respir J 2005; 26: 153–161. No. 2: Miller MR, Hankinson J, Brusasco V, et al. Standardisation of spirometry. Eur Respir J 2005; 26: 319–338. No. 3: Wanger J, Clausen JL, Coates A, et al. Standardisation of the measurement of lung volumes. Eur Respir J 2005; 26: 511–522. No. 4: MacIntyre N, Crapo RO, Viegi G, et al. Standardisation of the single-breath determination of carbon monoxide uptake in the lung. Eur Respir J 2005; 26: 720–735.

  • Received March 24, 2005.
  • Accepted April 5, 2005.


View Abstract