
The Society
More than 35,000 clinicians, scientists and allied health professionals from over 160 countries are part of the ERS community
Become an ERS member

The Global Lung Function Initiative (GLI) is a Clinical Research Collaboration (CRC), bringing together physiologists, respirologists, epidemiologists and statisticians to improve how lung function tests are interpreted.
The Global Lung Function Initiative (GLI) has collected respiratory function outcomes from researchers and health care professionals from around the world. To date, the GLI network has produced reference equations for spirometry, transfer factor for carbon monoxide and lung volumes. The GLI network manages the lung function data repository. In addition, the GLI data repository is available to researchers to answer novel research questions.
The GLI network administers the lung function data repository, whereas the data are securely stored by the European Respiratory Society (ERS). All data are managed and stored according to current international best practice.
Watch this interview with Prof. Graham Hall, explaining the importance of the GLI.
Learn more about the work of the GLI.
Access a range of tools relating to the work of the GLI project
Access a range of resources and publications relating to the GLI project
The GLI equations are comprised of an equation that requires an additional spline function from a look-up table. This spline function helps to improve the accuracy in which normal lung growth and decline are characterised and avoids misinterpretation of spirometry data. The GLI team have developed several tools to facilitate the use of these equations for research, education and training purposes. We have also worked closely with manufacturers to ensure timely installation in commercially available devices.
For an Individual Subject
To calculate z-scores, percent predicted or the lower limit of normal for an individual subject we have developed a desktop programme as well as an Excel programme. Both are easy to use, and simply require entering a subject’s age, height, sex, ethnic group and observed values.
To calculate z-scores, percent predicted or the lower limit of normal within your existing pulmonary function testing equipment, please contact your manufacturer and provide them with a link to this website.
Data Set
For existing data sets, or research studies we have developed several macros for different software to facilitate this. There is a desktop programme that will read .dat (tab delimited) files, an Excel macro embedded within a spreadsheet that simply requires you to copy/paste or input an existing spreadsheet. An SAS macro programme, as well as an R programme script will be uploaded shortly. All have been tested for accuracy, but potential bugs may exist. Please contact the analytical team if you incur any problems. The analytical team does not provide support for using these macros. These programmes are not yet compatible on Mac devices.
Source Code
Software developers or individual users can access the source code for programming the GLI equations. The source will be available on the website short, in the meantime you may contact the analytical team.
Measurements of lung function are essential components of the diagnosis and clinical care of individuals with pulmonary disease. Lung function is used beyond the care of those with chronic lung disease to inform clinical care decisions, such as eligibility for certain medications, surgeries and transplant. Pulmonary function tests (PFTs) are also used to assess eligibility for certain occupations and have been used to assess eligibility for disability and insurance benefits.
Historically, differences in lung function observed between populations have been attributed to the combination of genetic and environmental influences on the growth and development of the lungs. Consequently, the ATS and ERS previously recommended population-specific reference equations, often summarised as race/ethnicity specific reference equations. However, this approach likely underestimates the impact that social determinants have on lung health, and the use of race/ethnic specific equations may obscure true disparities in lung health. The reasons for observed differences in lung function between people around the world are multifactorial and not fully understood. There are ongoing efforts to elucidate the geographical, environmental, genetic, and social determinants of health that play a role in explaining these observed differences. Until high quality evidence is available to inform more accurate and precise interpretation of lung function, the ATS and ERS have recommended race-neutral approaches to interpret lung function measures. Irrespective of which reference equation is used, interpretation of lung function measurements requires careful consideration of an individual’s symptoms and medical history when used to make clinical, employment, and insurance decisions.
New approaches will be considered with a prioritised goal of ensuring health equity and with emphasis on considering lung function in the context of an individual’s medical history and clinical presentation.
Several manufacturers and their clientele have enquired about the lack of a predicted shape of the flow-volume curve and the normal scatter when using the GLI-2012 equations. Since data on PEF and other instantaneous flows were not collected for the GLI-2012, it is not possible for manufacturers to depict such predicted contours or a representative flow-volume curve using GLI-2012 prediction equations. The degree of imprecision that is inherent in producing such ‘ideal’ contours is so high that we recommend they should not be shown, focus instead being placed on the actual results. By presenting spirometry results in the form of a pictogram, the clinician, technician, physiologist and patient should all be able to interpret results far more accurately than previously, negating the need for any idealised (and often incorrect) flow-volume loop. We will, however, develop a set of curves with different diagnostic features to assist interpretation with respect to the shape of such curves, which will be posted on the website as part of the educational material currently being developed.
If, however, the manufacturer feels the need to continue to depict idealised curves, we suggest continuing to use prediction equations that were hitherto used.
Access this graphic to see clinical validation for GLI equations
The software/equations are published open access under CC-BY-NC licence.
Usage requires appropriate acknowledgement. It does not permit commercial usage. If you want to use the material commercially, please contact the ERS at permissions@ersj.org.uk.
Individuals over the age of 80 years are rarely included in research studies, therefore we have very little data to describe what ‘healthy’ lung function is. Some manufacturers extrapolate or combine reference equations to allow for interpretation of results in those over the age of 80 years. We recommend against this practice. For individuals over 80 years, the recommendation is to assume (and enter) an individuals age as 80 years into the system, and interpret the results with a greater degree of uncertainty. A note in the interpretation should be added that the confidence in interpretation is low.
GLI incurs a cost each time the API is used, and we need to ensure that we can sustain the demand moving forward. You can click here to find out how to get an API Key and how much it costs. You can scroll down to the Programmatic Access API section.
A z-score describes how far a measured value is from the range that we expect in otherwise healthy individuals. A z-score of zero represents the measured value as similar to the average of the reference population. A negative z-score represents values below the average, and a positive z-score represents values above the average. The lower limit of normal is defined to identify individuals whose measured values are outside the range we expect in healthy individuals. Since we are generally interested in people who have low lung function, we set the lower limit of normal at -1.645. At this level, approximately 5% of healthy individuals will be below the limit; a 5% false positive rate. In some circumstances, we may set the limit to -1.96. Only 2.5% of healthy individuals are below this limit, and there is a 2.5% false positive rate.
Historically, percent predicted has been used to describe how far an individual is from the average predicted value of a reference population, and 80% predicted has widely been used to define the lower limit of normal. This assumes that the variability of values around the predicted value is similar, and that 80% predicted is the lower limit of normal for people of all ages. We have now have observed that the range of values (biological variability) of lung function in younger children and older adults is much wider than for young adult males. Consequently, 80% predicted is biased by age, height and sex.
The GLI equations are comprised of an equation that requires an additional spline function from a look-up table. This spline function helps to improve the accuracy in which normal lung growth and decline are characterised and avoids misinterpretation.
The GLI initially published ethnic-specific reference equations based on available data and the available science at the time. Since then, there have been several scientific studies published that highlight potential biases in how we originally perceived the influence of race and ethnicity on lung health. The new evidence prompted the GLI Network to develop race-neutral equations as an alternative to race-ethnicity-specific equations. By assuming that non-white individuals have lower lung function because of biological differences, we may have systemic under-treatment and perpetuation of existing disparities in health outcomes https://pubmed.ncbi.nlm.nih.gov/34913855/
GLI has always acknowledged that race and ethnicity are social constructs, and we incorrectly used these variables as proxies for biological differences between populations. None of the GLI analyses accounted for an individual’s socioeconomic circumstances or the environmental influences on lung health.
A change to race-neutral equations does not imply that these equations are more accurate or precise. These acknowledge that there is a greater range of biological variability in lung function, and greater uncertainty in interpreting values. We are advocating for the respiratory community to acknowledge the imprecision and inherent variability of lung function in people across the world and for future research to be conducted to explain why we observe differences between populations.
Data on PEF and other instantaneous flows were not collected by many sites, and it is not possible to develop reference equations for these parameters.