Application of the European Respiratory Society/American Thoracic Society Spirometry Standards and Race-Neutral Equations in the COPDGene Study

Abstract

Rationale

For spirometry interpretation, the European Respiratory Society (ERS) and American Thoracic Society (ATS) recommend using z-scores, and the ATS has recommended using Global Lung Initiative (GLI) “Global” race-neutral reference equations. However, these recommendations have been variably implemented, and the impact has not been widely assessed in clinical or research settings.

Objectives

To evaluate the ERS/ATS classification of airflow obstruction severity.

Methods

In COPDGene (Genetic Epidemiology of COPD Study) (N = 10,108), airflow obstruction has been defined by an FEV₁/FVC ratio <0.70, with spirometric severity graded from class 1 to class 4 based on race-specific percent predicted (pp) FEV₁ cutoff points as recommended by the Global Initiative for Chronic Obstructive Lung Disease (GOLD). We compared the GOLD approach, using National Health and Nutrition Examination Survey III race-specific equations, versus the application of GLI Global equations using the ERS/ATS definition of airflow obstruction as an FEV₁/FVC ratio below the lower limit of normal and z-score–based FEV₁ cutoff points of –1.645, –2.5, and –4 (“zGLI Global”). We tested the four-tier severity scheme for association with chronic obstructive pulmonary disease outcomes.

Measurements and Main Results

The lowest agreement between ERS/ATS with zGLI Global and the GOLD classification was observed in individuals with milder disease (56.9% and 42.5% in GOLD stages 1 and 2, respectively), and race was a major determinant of redistribution. After adjustment for relevant covariates, zGLI Global distinguished all-cause mortality risk between normal spirometry and the first grade of chronic obstructive pulmonary disease (hazard ratio, 1.23; 95% confidence interval, 1.04–1.44; P = 0.014) and showed a linear increase in exacerbation rates with increasing disease severity in comparison with GOLD.

Conclusions

The zGLI Global severity classification outperformed the GOLD criteria in the discrimination of survival, exacerbations, and imaging characteristics.

At a Glance Commentary

Scientific Knowledge on the Subject

The European Respiratory Society (ERS) and the American Thoracic Society (ATS) recommend using z-scores and Global Lung Initiative “Global” race-neutral reference equations for spirometry interpretation. However, these recommendations have been variably implemented, and the impact has not been widely assessed in clinical or research settings.

What This Study Adds to the Field

In COPDGene (the Genetic Epidemiology of COPD Study), chronic obstructive pulmonary disease severity classification based on ERS/ATS 2021 standards and Global Lung Initiative Global reference equations could discriminate survival, imaging characteristics, and exacerbation rates better than the Global Initiative for Chronic Obstructive Lung Disease classification. Additionally, individuals with more modest disease severity were most commonly redistributed using ERS/ATS criteria, and race was a major determinant of subject redistribution.

The American Thoracic Society (ATS) has defined airway obstruction as an FEV₁/FVC ratio below the lower limit of normal (LLN) since 1991 (1). That definition continued through the 2005 Interpretation Statement, which was a joint ATS/European Respiratory Society (ERS) effort (2). However, for the past two decades, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) has defined chronic obstructive pulmonary disease (COPD) as a fixed FEV₁/FVC ratio of <0.7, and the reduction of percent predicted FEV₁ (ppFEV₁) has been used to assess severity (3). The GOLD standards have been accepted in everyday clinical practice and in the research field. Therefore, the majority of COPD studies have recruited and classified subjects using the GOLD criteria and have provided current scientific knowledge of COPD based on these criteria (4, 5).

In a document published online in 2021 (and in its final form in 2022), the ERS and ATS (6) proposed changes to the parameters and algorithms used in spirometry interpretation. To stage severity, the proposal used z-scores (the number of standard deviations by which the measured value is above or below the mean) rather than percent predicted values. Additionally, the use of the LLN and the upper limit of normal, usually set to the 5th and 95th percentiles of a population, respectively, was suggested. The use of a fixed FEV₁/FVC ratio <0.7 and ppFEV₁ to stage obstructive lung disease severity, as in the GOLD criteria, was strongly discouraged.

A subsequent official ATS statement (7) published in 2023 encouraged replacing race- and ethnicity-specific equations with race-neutral average reference equations, referred to as the Global Lung Initiative (GLI) Global criteria, developed by Bowerman and coworkers (8). The main rationale was that the social construct of race should not be used to infer biological characteristics, which perpetuates the false idea that race distinguishes people on the basis of innate and immutable features. Furthermore, normalization of differences with race-specific equations was seen as a potential contributor to medical harms by reducing the attention paid to other modifiable risk factors for reduced pulmonary function. For specific subsets of people (e.g., Black patients), this could result in delaying or missing disease diagnoses or hindering access to therapies. However, the 2023 ATS Statement indicates that “further study on the consequences of adopting the recommendations is imperative,” and the use of the GLI Global equations is still not fully implemented in clinical or research settings.

Therefore, we aimed to adopt the proposed recommendations to derive new severity classifications in a large COPD epidemiology study and to compare these published standards with the GOLD criteria, which have been previously used in many COPD studies. We tested the hypothesis that adhering to the 2021 ERS/ATS standards and the ATS 2023 recommendation on race-neutral reference values could improve diagnosis, subgroup classification, and prognostic stratification in COPD compared with the widely used approach of applying an FEV₁/FVC ratio <0.70 and GOLD staging classes 1–4 based on race-specific ppFEV₁ cutoff points.

Methods

Participants

COPDGene (The Genetic Epidemiology of COPD Study; NCT00608764) is a multicenter observational study (9). Self-identified non-Hispanic White and African American current and former smokers (⩾10 pack-year history) aged 45–80 years with and without COPD and healthy never-smoker controls were enrolled from 2008 to 2011 at 21 participating centers across the United States. See online supplement for details of participant assessments. All participants provided written informed consent. The COPDGene study was approved by the institutional review boards at all participating centers.

We analyzed current and former smokers with available postbronchodilator spirometry at the baseline visit (N = 10,108). Classifications are shown in Table 1. For “GOLD-NHANES,” airflow obstruction was defined by a postbronchodilator FEV₁/FVC ratio <0.70. National Health and Nutrition Examination Survey (NHANES) III race-specific equations were used to calculate ppFEV₁ for assessing the severity of obstruction based on GOLD stages 1–4. Normal spirometry was considered to represent GOLD stage 0. The preserved-ratio impaired spirometry (PRISm) category included participants with an FEV₁/FVC ratio ⩾0.70 and ppFEV₁ <80.

Table 1.

Classification Criteria

GOLD-NHANES		zGLI Race-specific		zGLI Global
GOLD 0	FEV1/FVC ⩾0.70, ppFEV1 ⩾80	Pre-COPD	zFEV1/FVC ⩾–1.645, zFEV1 ⩾–1.645	Pre-COPD	zFEV1/FVC ⩾–1.645, zFEV1 ⩾–1.645
GOLD 1	FEV1/FVC <0.70, ppFEV1 ⩾80	Borderline	zFEV1/FVC <–1.645, zFEV1 ⩾–1.645	Borderline	zFEV1/FVC <–1.645, zFEV1 ⩾–1.645
GOLD 2	FEV1/FVC <0.70, 50 ⩽ ppFEV1 < 80	Mild	zFEV1/FVC <–1.645, −2.5 ⩽ zFEV1 < –1.645	Mild	zFEV1/FVC <–1.645, −2.5 ⩽ zFEV1 < –1.645
GOLD 3	FEV1/FVC <0.70, 30 ⩽ ppFEV1 < 50	Moderate	zFEV1/FVC <–1.645, −4 ⩽ zFEV1 <–2.5	Moderate	zFEV1/FVC <–1.645, −4 ⩽ zFEV1 < –2.5
GOLD 4	FEV1/FVC <0.70, ppFEV1 <30	Severe	zFEV1/FVC <–1.645, zFEV1 <–4	Severe	zFEV1/FVC <–1.645, zFEV1 <–4
PRISm	FEV1/FVC ⩾0.70, ppFEV1 <80	PRISm	zFEV1/FVC ⩾–1.645, zFEV1 <–1.645	PRISm	zFEV1/FVC ⩾–1.645, zFEV1 <–1.645

Definition of abbreviations: COPD = chronic obstructive pulmonary disease; GOLD = Global Initiative for Chronic Obstructive Lung Disease; NHANES = National Health and Nutrition Examination Survey; pp = percent predicted; PRISm = preserved-ratio impaired spirometry; zGLI = z-score–based Global Lung Function Initiative classification.

“GOLD-NHANES” is NHANES III race-specific, and cutpoints are based on percent predicted. “zGLI Race” and “zGLI Global” each apply the z-score approach for severity classification.

According to the 2021 ERS/ATS standards, airflow obstruction was defined by a postbronchodilator FEV₁/FVC ratio below the LLN. For subjects with obstruction, severity classes were based on GLI z-score–based FEV₁ and were labeled as borderline, mild, moderate, and severe, analogous to GOLD stages 1–4. Two different classifications, zGLI Race-Specific and zGLI Global, were derived based on GLI race-specific and GLI Global race-neutral reference equations, respectively, which are implemented in the rspiro package for R (version 0.4). The class names follow the ERS/ATS 2021 lung function impairment scaling system except for the “borderline” group, which corresponds to a normal FEV₁ with a reduced FEV₁/FVC ratio. Normal spirometry findings were described as “pre-COPD.” The PRISm category included subjects with an FEV₁/FVC ratio of at least the LLN and a FEV₁ below the LLN as previously reported (10).

Statistical Analysis

We compared the three abovementioned severity classifications with the primary outcome of all-cause mortality. The Bangdiwala approach (11) was used to assess agreement between the severity classifications descriptively through agreement charts and quantitatively through the weighted B-statistic, which adjusts for the frequency of each severity class. The B-statistic is a proportion, so it ranges from 0 for no agreement to 1 for perfect agreement. Interpretation is as follows: values from 1 to 0.65 indicate almost perfect agreement, those from 0.65 to 0.35 suggest substantial agreement, those from 0.35 to 0.15 suggest moderate agreement, those from 0.15 to 0.05 suggest fair agreement, and those below 0.05 suggest poor agreement (12).

Cox proportional hazards models for all-cause mortality were fitted with adjustment for relevant covariates. Concordance was used to measure the degree to which the model correctly predicts the order of events (e.g., death). On a scale from 0 to 1, higher concordance indicates stronger performance, with a value of 0.5 implying no discrimination beyond chance. Subjects classified as GOLD stage 0, zGLI Race-Specific pre-COPD, and zGLI Global pre-COPD were used as the reference groups for GOLD-NHANES, zGLI Race-Specific, and zGLI Global classifications, respectively.

Generalized linear regression models were used to test associations between the severity classes and continuous clinical (St. George Respiratory Questionnaire [SGRQ]; Modified Medical Research Council [mMRC] dyspnea scale; body mass index, airflow obstruction, dyspnea, and exercise capacity [BODE] index; 6-minute-walk test [6MWT] distance) and chest computed tomography (CT) imaging characteristics (percent emphysema, percent air trapping, and the square root of the wall area of a hypothetical airway with a 10-mm internal perimeter, used to quantify airway wall thickness [Pi10]). Negative binomial regression models were used to test the association between severity classes and prospective exacerbation frequency. All models were adjusted for age, sex, race, body mass index, current smoking, and pack-years of smoking. For exacerbations recorded during the study, models were additionally adjusted for the history of exacerbations in the 1 year before enrollment. All analyses were performed using R version 4.2.0. A two-sided α-value of 0.05 was considered statistically significant.

Results

Agreement between GOLD-NHANES and zGLI Global Classes

Participant characteristics are provided in Table E1 in the online supplement. Briefly, of the 10,108 subjects included, 53.4% (n = 5,394) were male and 46.6% (n = 4714) were female. Proportions of non-Hispanic White and African American individuals were 66.9% (n = 6,760) and 33.1% (n = 3,348), respectively.

The distribution of COPDGene participants by disease severity according to zGLI Global class is shown in Figure 1. Agreement between the two classifications is visually displayed by a Bangdiwala plot in Figure 2A, a Sankey diagram in Figure 2B, and a redistribution table in Table E2. The weighted Bangdiwala B-statistic of 0.84 indicates almost perfect agreement between GOLD-NHANES and zGLI Global classes. The major redistributions occurred from GOLD stage 2 to less severe borderline (n = 413, 21.4%) and more severe moderate (n = 396, 20.6%) classifications, from GOLD stage 1 to pre-COPD (n = 315, 40.0%), and from GOLD stage 4 to the less severe moderate category (n = 147, 24.3%).

Figure 1.

Distribution of COPDGene (Genetic Epidemiology of COPD Study) subjects by zGLI Global Classification. COPD = chronic obstructive pulmonary disease; GLI = Global Lung Function Initiative; PRISm = preserved-ratio impaired spirometry; zGLI = z-score–based Global Lung Function Initiative classification.

Figure 2.

Bangdiwala agreement plot (A) and Sankey diagram (B) between Global Initiative for Chronic Obstructive Lung Disease (GOLD)–National Health and Nutrition Examination Survey (NHANES) and z-score–based Global Lung Function Initiative (zGLI) Global severity classes. (A) The agreement plot is a k × k square, where k is the total sample size (N = 10,108). The labels on the top and right axes correspond to the GOLD and GLI Global severity classes, respectively. Within each larger rectangle (maximum possible agreement), each black, gray, and white rectangle indicates progressively lower agreement, as displayed in the agreement legend. In the case of perfect agreement, all the rectangles would be black squares with a 45° diagonal line touching the corners of each square. The weighted Bangdiwala (“B”) statistic assesses agreement between multiple classes, adjusting for the frequency of each severity class. (B) The Sankey diagram visually describes the redistribution of participants from GOLD–NHANES to zGLI Global classification. The major redistributions occurred from GOLD stage 2 to less severe borderline (n = 413, 21.4%) and to more severe moderate (n = 396, 20.6%) categories, from GOLD stage 1 chronic obstructive pulmonary disease to pre–COPD (n = 315, 40.0%), and from GOLD stage 4 to the less severe moderate category (n = 147, 24.3%). See Table E2 for the numbers of patients and percentages for each class. PR = preserved-ratio impaired spirometry.

GOLD Stage 2 Redistribution

The GOLD stage 2 class was redistributed over five different zGLI Global severity categories and showed the lowest agreement rate (42.5%). To further categorize the intraclass heterogeneity, participants with GOLD stage 2 obstruction were regrouped as “GOLD2-down” (n = 553, 31.3%) if their obstruction was “downstaged” to a less severe category, zGLI Global pre-COPD or borderline; “GOLD2-stable” (n = 818, 46.3%) if they remained in the corresponding zGLI Global mild class; and “GOLD2-up” (n = 396, 22.4%) if they were “upstaged” to the more severe zGLI Global moderate category. Patients reclassified in zGLI PRISm classes were not included in any of the redistribution analyses. The GOLD2-up subjects were significantly younger, taller, and more likely to be men, African American, and current smokers (Table 2). Compared with the GOLD2-stable group, their SGRQ score was higher, suggesting worse disease-related quality of life, with a difference greater than the minimum clinically important difference of 4 points, along with higher exacerbation rates and mMRC and BODE scores. These subjects also showed a higher square root of the wall area of a hypothetical airway with a 10-mm internal perimeter and more emphysema and gas trapping on CT compared with the other classes.

Table 2.

Demographic, Clinical, and Imaging Characteristics in GOLD Stage 2 Reclassified by zGLI Global Criteria

	GOLD2-Down (n = 553; 31.3%)*	GOLD2-Stable (n = 818; 46.3%)	GOLD2-Up (n = 396; 22.4%)†	P Value
Baseline demographic characteristics
Age, yr	63.6 ± 8.3	62.3 ± 9.1	60.4 ± 9.2	<0.001
Height, cm	169.8 ± 9.1	168.8 ± 9.3	170.9 ± 9.7	0.001
Weight, kg	82.3 ± 18.2	81.6 ± 19.1	84.7 ± 20.2	0.023
BMI, kg/m2	28.5 ± 5.7	28.6 ± 6.2	29.0 ± 6.4	0.44
Sex				<0.001
Male	272 (49.2%)	385 (47.1%)	256 (64.6%)
Female	281 (50.8%)	433 (52.9%)	140 (35.4%)
Self-identified race				<0.001
Non-Hispanic White	551 (99.6%)	616 (75.3%)	177 (44.7%)
African American	2 (0.4%)	202 (24.7%)	219 (55.3%)
Smoking status				0.015
Former smoker	301 (54.4%)	400 (48.9%)	179 (45.2%)
Current smoker	252 (45.6%)	418 (51.1%)	217 (54.8%)
Smoking, pack-years	51.0 ± 26.4	51.4 ± 27.1	49.2 ± 27.8	0.39
Clinical characteristics
Postbronchodilator FEV1, L	2.2 ± 0.5	1.8 ± 0.5	1.6 ± 0.4	<0.001
Postbronchodilator ppFEV1
NHANES	73.8 ± 3.4	63.8 ± 6.9	55.7 ± 4.8	<0.001
GLI Global	78.0 ± 3.8	64.4 ± 5.2	53.5 ± 4.2	<0.001
SGRQ total score	24.6 ± 19.5	34.7 ± 20.7	41.2 ± 21.9	<0.001
mMRC score	1.1 ± 1.3	1.7 ± 1.4	2.1 ± 1.4	<0.001
6MWT distance, m	436.0 ± 101.5	395.6 ± 106.6	358.2 ± 114.3	<0.001
BODE index	0.9 ± 1.2	2.1 ± 1.6	3.0 ± 1.7	<0.001
Exacerbations in 1 yr before enrollment	0.38 ± 0.90	0.59 ± 1.08	0.72 ± 1.19	<0.001
⩾1 severe exacerbation in 1 yr before enrollment				<0.001
Yes	52 (9.4%)	130 (15.9%)	100 (25.3%)
No	501 (90.6%)	688 (84.1%)	296 (74.7%)
Imaging characteristics
Emphysema, %	5.7 ± 6.1	8.5 ± 8.8	9.0 ± 9.0	<0.001
Gas trapping, %	22.9 ± 12.5	28.7 ± 15.6	32.3 ± 17.1	<0.001
Pi10, mm	2.4 ± 0.5	2.7 ± 0.6	2.8 ± 0.6	<0.001

Definition of abbreviations: 6MWT = 6-minute-walk test; BMI = body mass index; GOLD = Global Initiative for Chronic Obstructive Lung Disease; mMRC = modified Medical Research Council; NHANES = National Health and Nutrition Examination Survey; Pi10 = square root of the wall area of a hypothetical airway with a 10-mm internal perimeter; pp = percent predicted; SGRQ = St. George Respiratory Questionnaire; zGLI = z-score–based Global Lung Function Initiative classification.

Continuous data presented as mean ± standard deviation.

^*Downstaged to a less severe zGLI Global category (pre-COPD or borderline).

^† Upstaged to the more severe zGLI Global moderate category.

The racial distribution was different across the three groups. The GOLD2-down group included only two African American subjects (0.4%), whereas African American subjects represented a quarter of the GOLD2-stable group (n = 202, 24.7%) and more than half of the GOLD2-up group (n = 219, 55.3%). Figure 3A shows that nearly all African American patients with GOLD stage 2 obstruction were stable as zGLI Global mild or upstaged to zGLI Global moderate. White subjects, on the contrary, were largely stable (n = 616, 47.8%) or downstaged (n = 551, 41.0%), with a smaller proportion (n = 177, 13.2%) upstaged to the zGLI Global moderate category.

Figure 3.

Racial distribution of Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage 2 (A) and GOLD stage 1 (B) subjects reclassified by z-score–based Global Lung Function Initiative (zGLI) Global. (A) GOLD stage 2 subjects were re-grouped as “GOLD2-down” (n = 553, 31.3%) when classified as zGLI Global pre–chronic obstructive pulmonary disease (COPD) (n = 140, 7.9%) or borderline (n = 413, 23.4%), “GOLD2-stable” (n = 818, 46.3%) when classified as zGLI Global mild, and “GOLD2-up” (n = 396, 22.4%) when classified as zGLI Global moderate. (B) GOLD stage 1 subjects were regrouped as “GOLD1-pre” (n = 315, 40.2%) when classified as zGLI Global pre-COPD and as “GOLD1-COPD” (n = 468, 59.8%) when classified as zGLI Global borderline (n = 448, 57.2%) or mild (n = 20, 2.6%).

Similar results were found in subjects with GOLD stage 1 obstruction (n = 787) (Table 3), who also showed a modest agreement rate with the zGLI Global borderline class (56.9%). These subjects were reclassified as “GOLD1-pre” (n = 315, 40.2%) if they were categorized as zGLI Global pre-COPD and as “GOLD1-COPD” (n = 468, 59.8%) if they were categorized as zGLI Global borderline or mild. Figure 3B illustrates the shift of African American individuals toward the more severe disease categories.

Table 3.

Demographic, Clinical, and Imaging Characteristics in and GOLD Stage 1 Reclassified by zGLI Global Criteria

	GOLD1 Pre-COPD (n = 315; 40.2%)	GOLD1 COPD (n = 468; 59.8%)	P Value
Baseline demographic characteristics
Age, yr	65.3 ± 7.8	59.4 ± 9.0	<0.001
Height, cm	172.9 ± 9.3	168.3 ± 9.6	<0.001
Weight, kg	82.5 ± 16.1	75.7 ± 15.7	<0.001
BMI, kg/m2	27.5 ± 4.6	26.7 ± 5.2	0.023
Sex			<0.001
Male	222 (70.5%)	228 (48.7%)
Female	93 (29.5%)	240 (51.3%)
Self-identified race			<0.001
Non-Hispanic White	267 (84.8%)	342 (73.1%)
African American	48 (15.2%)	126 (26.9%)
Smoking status			<0.001
Former smoker	170 (54.0%)	177 (37.8%)
Current smoker	145 (46.0%)	291 (62.2%)
Smoking, pack-years	47.7 ± 26.5	43.2 ± 23.2	0.012
Clinical characteristics
Postbronchodilator FEV1, L	2.8 ± 0.7	2.6 ± 0.6	<0.001
Postbronchodilator ppFEV1
NHANES	92.0 ± 9.3	89.9 ± 8.4	<0.001
GLI Global	94.8 ± 10.7	90.8 ± 10.5	<0.001
SGRQ total score	14.8 ± 15.1	20.3 ± 18.8	<0.001
mMRC score	0.6 ± 1.1	0.9 ± 1.2	0.002
6MWT distance, m	461.5 ± 108.4	456.0 ± 101.9	0.48
BODE index	0.5 ± 1.0	0.7 ± 1.1	0.023
Exacerbations in 1 yr before enrollment	0.14 ± 0.61	0.25 ± 0.74	0.036
⩾1 severe exacerbation in 1 yr before enrollment			0.79
Yes	14 (4.4%)	24 (5.1%)
No	301 (95.6%)	444 (94.9%)
Imaging characteristics
Emphysema, %	5.2 ± 4.9	5.6 ± 6.4	0.37
Gas trapping, %	19.5 ± 10.3	19.4 ± 12.3	0.89
Pi10, mm	2.1 ±  0.41	2.2 ± 0.5	<0.001

Definition of abbreviations: 6MWT = 6-minute-walk test; BMI = body mass index; GOLD = Global Initiative for Chronic Obstructive Lung Disease; mMRC = modified Medical Research Council; NHANES = National Health and Nutrition Examination Survey; Pi10 = square root of the wall area of a hypothetical airway with a 10-mm internal perimeter; pp = percent predicted; SGRQ = St. George Respiratory Questionnaire; zGLI = z-score–based Global Lung Function Initiative classification.

Continuous data presented as mean ± standard deviation.

We then tested whether the shift from percent-predicted values to z-scores or the shift from race-specific to race-neutral equations was the major contributor to these trends. We separately analyzed the GOLD stage 2 redistribution by zGLI Race-Specific classes (step 1) and, subsequently, the shift from zGLI Race-Specific to zGLI Global classification (step 2). Table E2 summarizes the agreement between classes in step 1 and step 2. In step 1, we found a difference by race among the three groups (P = 0.01), but the proportions of African American individuals in the GOLD2-down and GOLD2-up groups were relatively similar (24.9% and 22.1%, respectively; Figure 4A and Table 4). In step 2, on the contrary, the division was more drastic: 100% (n = 369) of the downstaged group was White, and 95.2% (n = 542) of the upstaged group was African American (P < 0.0001; Figure 4B and Table 5).

Figure 4.

Racial distribution of (A) Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage 2 subjects by z-score–based Global Lung Function Initiative (zGLI) Race-Specific (step 1) and (B) subjects with zGLI Race-Specific mild obstruction by zGLI Global (step 2). (A) In step 1, GOLD stage 2 subjects were regrouped as “GOLD2-down” (n = 369, 21.6%) when classified as zGLI Race-Specific pre-COPD (n = 135, 7.9%) or borderline (n = 234, 13.7%), “GOLD2-stable” (n = 796, 46.5%) when classified as zGLI Race-Specific mild, and “GOLD2-up” (n = 547, 31.9%) when classified as zGLI Race-Specific moderate. (B) In step 2, subjects with zGLI Race-Specific mild obstruction were regrouped as “zGLI Race-Specific mild-down” (n = 218, 27.4%) when classified as zGLI Global borderline, “zGLI Race-Specific mild-stable” when classified as zGLI Global mild (n = 474, 59.5%), and as “zGLI Race-Specific mild-up” when classified as zGLI Global moderate (n = 104, 13.1%).

Table 4.

Demographic, Clinical, and Imaging Characteristics in GOLD Stage 2 Subjects by zGLI Race-Specific Criteria

	GOLD2-Down (n = 369; 21.6%)*	GOLD2-Stable (n = 796; 46.5%)	GOLD2-Up (n = 547; 31.9%)†	P Value
Baseline demographic characteristics
Age, yr	64.7 ± 8.2	62.0 ± 9.2	60.8 ± 8.7	<0.001
Height, cm	171.7 ± 8.8	169.9 ± 9.5	167.9 ± 9.3	<0.001
Weight, kg	82.7 ± 17.5	81.5 ± 18.2	82.6 ± 20.6	0.51
BMI, kg/m2	28.1 ± 5.7	28.2 ± 5.9	29.2 ± 6.5	0.005
Sex				0.003
Male	203 (55.0%)	431 (54.1%)	249 (45.5%)
Female	166 (45.0%)	365 (45.9%)	298 (54.5%)
Self-identified race				0.010
Non-Hispanic White	277 (75.1%)	562 (70.6%)	426 (77.9%)
African American	92 (24.9%)	234 (29.4%)	121 (22.1%)
Smoking status				0.018
Former smoker	197 (53.4%)	364 (45.7%)	284 (51.9%)
Current smoker	172 (46.6%)	432 (54.3%)	263 (48.1%)
Smoking, pack-years	49.7 ± 28.4	49.8 ± 25.5	51.7 ± 27.9	0.38
Clinical characteristics
Postbronchodilator FEV1, L	2.2 ± 0.5	1.9 ± 0.5	1.6 ± 0.4	<0.001
Postbronchodilator ppFEV1
NHANES	75.9 ± 2.9	66.7 ± 5.5	55.5 ± 3.9	<0.001
GLI Race-Specific	76.9 ± 2.8	66.9 ± 5.3	55.5 ± 4.2	<0.001
SGRQ total score	23.4 ± 20.0	32.7 ± 20.8	40.5 ± 20.8	<0.001
mMRC score	1.1 ± 1.3	1.6 ± 1.4	2.1 ± 1.3	<0.001
6MWT distance, m	422.4 ± 105.2	400.6 ± 111.4	380.1 ± 110.4	<0.001
BODE index	1.0 ± 1.4	1.8 ± 1.6	2.8 ± 1.6	<0.001
Exacerbations in 1 yr before enrollment	0.33 ± 0.83	0.52 ± 1.05	0.77 ± 1.20	<0.001
⩾1 severe exacerbation in 1 yr before enrollment				<0.001
Yes	30 (8.1%)	122 (15.3%)	123 (22.5%)
No	339 (91.9%)	674 (84.7%)	424 (77.5%)
Imaging characteristics
Emphysema, %	5.6 ± 5.9	7.8 ± 8.2	9.3 ± 9.2	<0.001
Gas trapping, %	23.5 ± 12.3	27.9 ± 15.9	30.9 ± 16.2	<0.001
Pi10, mm	2.3 ± 0.5	2.6 ± 0.6	2.8 ± 0.6	<0.001

Definition of abbreviations: 6MWT = 6-minute-walk test; BMI = body mass index; GOLD = Global Initiative for Chronic Obstructive Lung Disease; mMRC = modified Medical Research Council; NHANES = National Health and Nutrition Examination Survey; Pi10 = square root of the wall area of a hypothetical airway with a 10-mm internal perimeter; pp = percent predicted; SGRQ = St. George Respiratory Questionnaire; zGLI = z-score–based Global Lung Function Initiative classification.

Continuous data presented as mean ± standard deviation.

^*Downstaged to a less severe zGLI Global category (pre-COPD or borderline).

^† Upstaged to the more severe zGLI Global moderate category.

Table 5.

Demographic, Clinical, and Imaging Characteristics in zGLI Race-Specific Mild Obstruction by zGLI Global Criteria

	zGLI-Race Specific Down (n = 218; 27.4%)*	zGLI-Race Specific Stable (n = 474; 59.5%)	zGLI-Race Specific Up (n = 104; 13.1%)†	P Value
Baseline demographic characteristics
Age, yr	60.1 ± 7.9	63.4 ± 9.5	59.7 ± 9.4	<0.001
Height, cm	168.3 ± 9.1	169.7 ± 9.6	173.8 ± 9.0	<0.001
Weight, kg	79.3 ± 16.9	82.1 ± 18.4	83.3 ± 19.9	0.091
BMI, kg/m2	28.0 ± 5.4	28.5 ± 6.0	27.6 ± 6.4	0.27
Sex				<0.001
Male	101 (46.3%)	250 (52.7%)	78 (75.0%)
Female	117 (53.7%)	224 (47.3%)	26 (25.0%)
Self-identified race				<0.001
Non-Hispanic White	218 (100.0%)	339 (71.5%)	5 (4.8%)
African American	0 (0.0%)	135 (28.5%)	99 (95.2%)
Smoking status				<0.001
Former smoker	97 (44.5%)	241 (50.8%)	27 (26.0%)
Current smoker	121 (55.5%)	233 (49.2%)	77 (74.0%)
Smoking, pack-years	50.7 ± 22.4	51.5 ± 27.4	39.6 ± 20.2	<0.001
Clinical characteristics
Postbronchodilator FEV1, L	2.1 ± 0.5	1.9 ± 0.5	1.8 ± 0.4	<0.001
Postbronchodilator ppFEV1
GLI Race-Specific	72.3 ± 2.4	65.6 ± 4.5	61.4 ± 4.2	<0.001
GLI Global	75.9 ± 2.9	65.7 ± 4.7	55.6 ± 3.7	<0.001
SGRQ total score	30.2 ± 20.8	32.4 ± 20.1	39.9 ± 23.2	<0.001
mMRC score	1.3 ± 1.3	1.6 ± 1.4	1.9 ± 1.4	0.001
6MWT distance, m	436.7 ± 99.4	396.7 ± 108.4	338.8 ± 121.7	<0.001
BODE index	1.0 ± 1.2	1.9 ± 1.6	3.0 ± 1.9	<0.001
Exacerbations in 1 yr before enrollment	0.52 ± 1.12	0.51 ± 1.01	0.54 ± 1.07	0.97
⩾1 severe exacerbation in 1 yr before enrollment				0.014
Yes	29 (13.3%)	68 (14.3%)	26 (25.0%)
No	189 (86.7%)	406 (85.7%)	78 (75.0%)
Imaging characteristics
Emphysema, %	6.0 ± 6.4	8.8 ± 9.1	7.4 ± 6.8	<0.001
Gas trapping, %	22.9 ± 13.5	29.3 ± 15.7	34.0 ± 18.5	<0.001
Pi10, mm	2.5 ± 0.5	2.6 ± 0.6	2.8 ± 0.6	<0.001

Definition of abbreviations: 6MWT = 6-minute-walk test; BMI = body mass index; GOLD = Global Initiative for Chronic Obstructive Lung Disease; mMRC = modified Medical Research Council; NHANES = National Health and Nutrition Examination Survey; Pi10 = square root of the wall area of a hypothetical airway with a 10-mm internal perimeter; pp = percent predicted; SGRQ = St. George Respiratory Questionnaire; zGLI = z-score–based Global Lung Function Initiative classification.

Continuous data presented as mean ± standard deviation.

^*Downstaged to a less severe zGLI Global category (pre-COPD or borderline).

^† Upstaged to the more severe zGLI Global moderate category.

Survival Analysis

During a median follow-up time of 10.5 years ± 4.3 (standard deviation), 2,676 (26.5%) participants died. Kaplan-Meier survival curves for all-cause mortality by GOLD-NHANES, zGLI Race-Specific, and zGLI Global classifications are shown in Figure E1. Unadjusted Cox proportional hazards models (Table E3) showed similar concordances for GOLD-NHANES, zGLI Race-Specific, and zGLI Global (0.70, 0.69 and 0.69, respectively). After adjustment for age, sex, race, height, and smoking status, concordance was 0.73 for all three classifications. However, as displayed in Table 6, only the zGLI Global model provided discrimination between pre-COPD and borderline COPD groups in terms of mortality (hazard ratio, 1.23; 95% confidence interval, 1.04–1.44; P = 0.014).

Table 6.

Comparison of All-Cause Mortality by Severity Class

Adjusted HR* (95% CI)
	GOLD-NHANES		zGLI Race-specific	zGLI Global
GOLD 0	Ref	Pre-COPD	Ref	Ref
GOLD 1	1.03 (0.85–1.24)	Borderline	1.13 (0.94–1.37)	1.23† (1.04–1.44)
GOLD 2	2.11‡ (1.88–2.37)	Mild	1.99‡ (1.73–2.30)	1.95‡ (1.70–2.25)
GOLD 3	4.11‡ (3.65–4.62)	Moderate	3.43‡ (3.10–3.80)	3.77‡ (3.42–4.17)
GOLD 4	9.52‡ (8.37–10.8)	Severe	8.37‡ (7.42–9.43)	8.35‡ (7.34–9.50)
PRISm	1.88‡ (1.63–2.17)	PRISm	1.94‡ (1.68–2.22)	1.83‡ (1.58–2.12)

Definition of abbreviations: CI = confidence interval; COPD = chronic obstructive pulmonary disease; GOLD = Global Initiative for Chronic Obstructive Lung Disease; NHANES = National Health and Nutrition Examination Survey; PRISm = preserved-ratio impaired spirometry; zGLI = z-score–based Global Lung Function Initiative classification.

“GOLD-NHANES” is NHANES III race-specific and cutpoints are based on percent predicted. “zGLI Race-Specific” and “zGLI Global” each apply z-score approach for severity classification.

^*Cox proportional hazard model adjusted for age, sex, race, height, and smoking status.

^† P < 0.05.

^‡ P < 0.001.

Clinical and Imaging Characteristics by Severity Class

Table E4 shows the comparison of clinical and imaging characteristics by severity class using adjusted linear models. All three classifications provided discrimination between normal spirometry and the first class of disease (GOLD stage 1 for GOLD-NHANES and borderline for the zGLI classes) for SGRQ and mMRC scores, whereas they failed for 6MWT distance and BODE score. GOLD-NHANES staging did not yield a difference in CT-detected gas trapping between PRISm and GOLD stage 0. The zGLI classifications, on the contrary, always offered separation between PRISm and pre-COPD classifications.

Exacerbations by Severity Classes

Table E5 shows a comparison of exacerbation rates by severity class using adjusted negative binomial models. Incidence rate ratios (IRRs) for the first class of disease compared with the normal spirometry class were significantly different, except for exacerbations requiring hospitalization in the 12 months before enrollment by GOLD criteria (IRR, 1.1; 95% confidence interval, 0.6–1.9; P = 0.76). Figure 5 displays IRRs by severity classifications. zGLI Global is the only classification scheme that shows a clear increasing trend in exacerbation rates across the borderline to severe classes for all of the considered variables.

Figure 5.

Exacerbation rates by severity class. Incidence rate ratios are displayed on the y-axis and are adjusted for age, sex, race, height, smoking status, and history of exacerbations in the 1 year before enrollment. “GOLD-NHANES” is NHANES III race-specific, and cutoff points are based on percent predicted. “zGLI Race-Specific” and “zGLI Global” each apply a z-score approach for severity classification. BL = borderline; GLI = Global Lung Function Initiative; GOLD = Global Initiative for Chronic Obstructive Lung Disease; NHANES = National Health and Nutrition Examination Survey; P1 = baseline visit; P2 = 5-year follow-up visit; zGLI = z-score–based Global Lung Function Initiative classification.

Discussion

In a large multicenter study of smokers, we demonstrated that a severity classification based on the 2021 ERS/ATS standards and the GLI Global reference equations differentiates prognoses, imaging parameters, and exacerbation rates better than the GOLD criteria. The zGLI Global classification provided discrimination for all-cause mortality risk between “normal” subjects and the first grade of COPD, separated PRISm from normality for all the considered variables, and showed a clear increase in exacerbation rates with increasing obstruction severity. Additionally, our analysis showed that individuals with more modest disease severity (GOLD stages 1 and 2) were most commonly redistributed using ERS/ATS criteria and that race was a major determinant of subject redistribution.

Several publications in the past few years have supported the ERS/ATS recommendations. Swanney and coworkers (13) documented a wide variability in the LLN of FEV₁/FVC ratio, with a marked negative age dependency. The use of a fixed threshold progressively increased the percentage of healthy subjects being labeled as having obstruction as age increased, reflecting the widening gap between the LLN and the 0.70 FEV₁/FVC ratio cutoff point. Miller and coworkers (14) confirmed these findings by comparing the fifth percentile as the LLN versus the GOLD fixed cutoffs of FEV₁/FVC ratio <0.7 and ppFEV <80. In 11,413 patients, the GOLD criteria misclassified 24% of subjects, giving false-positive classification for airflow obstruction to significantly more men and older patients while missing the diagnosis in younger patients.

In regard to staging, several z-score cutoffs have been proposed in recent years, a summary of which can be found in Table 7. Desbordes and coworkers (15) recently studied the possible clinical implication of 2021 ERS/ATS standards on a cohort of 1,323 patients between 2005 and 2018 at UZ Leuven in Belgium and found that older patients with low FEV₁/FVC ratios were reclassified as having milder disease by the new standards because the FEV₁ values were closer to normality.

Table 7.

Comparison between z-Score–based Cutoffs Proposed in Literature for Staging Severity

Study	Proposed Cutoffs
Quanjer et al. (26), 2014	LLN for FEV1/FVC; z-scores of –2, –2.5, –3, and –4 for FEV1 to stage severity of airflow limitation
Miller et al. (27), 2021	Four-tier grading based on z-scores of –1.645, –3, and –5 proposed for grading spirometry and TLCO using large cohort (N = 13,829) and survival data
Stanojevic et al. (6), 2022	ERS/ATS 2021 proposed three-level (four-tier) severity system that considered z-scores >–1.645 as normal, between –1.645 and –2.5 as mild, between –2.5 and –4 as moderate, and <–4.0 as severe

Definition of abbreviations: ATS = American Thoracic Society; ERS = European Respiratory Society; LLN = lower limit of normality; TLCO = single-breath transfer factor of the lungs for carbon monoxide.

Increasing evidence from the literature challenges the use of race-specific spirometry equations. Baugh and coworkers (16) compared the use of race-specific NHANES III and GLI “other” (race-neutral) equations in SPIROMICS (Subpopulations and Intermediate Outcome Measures in COPD Study), a multicenter study of ever-smokers with and without COPD. African American subjects appeared to have better lung function than non-Hispanic White individuals (ppFEV₁ 76.8% vs. 71.8%) with race-specific equations, but this was reversed using the GLI other equations (ppFEV₁ 70.0% vs. 77.9%). McCormack and coworkers (17) conducted a post hoc analysis of NHANES III data and demonstrated that mortality between White and Black individuals was more similar if a race-neutral approach was used rather than a race-specific one. Such findings supported the idea that race-specific equations may underestimate COPD severity in Black participants. Bowerman and coworkers (8) found systematic differences in the interpretation of pulmonary function test results in the NHANES III data with the use of the race- and ethnicity-specific NHANES III equations and sex-specific, race-neutral NHANES III equations that included Cormic index (ratio of sitting height to standing height), a proxy for chest size. Under the first approach, non-Hispanic Black individuals were more likely to have their function interpreted as better, whereas the opposite trend was observed in non-Hispanic White and Mexican American individuals.

More recent papers in COPDGene addressed the problem of severity classification in COPD with particular regard to race and reference equations. Two papers from Regan and coworkers highlighted the risk of COPD misdiagnosis in African American subjects when using the GOLD criteria. In the first (18), it was demonstrated that the fixed ratio of FEV₁/FVC <0.7 underdiagnosed potential COPD in African American individuals compared with broader diagnostic criteria. This could be explained by disproportionate reductions in FVC relative to FEV₁ leading to a higher ratio and being associated with social deprivation based on the Area Deprivation Index, which was found to be higher in the African American group.

In the second paper (19), the use of “race-reversed” equations (non-Hispanic White–specific equations for African American subjects and vice versa) reclassified 774 (43%) African American subjects from GOLD stage 0 to PRISm. In each GOLD stage, African American participants were younger, had fewer smoking pack-years and more current smoking, but had more exacerbations, lower 6MWT distance, greater dyspnea, and worse BODE and SGRQ scores. Bhatt and coworkers (20) proposed a new severity classification scheme (Staging of Airflow obstruction by Ratio) based on the FEV₁/FVC ratio that provided discrimination for mortality similar to the GOLD classification, but with a more uniform gradation of disease severity, which was less sensitive to race and other demographic characteristics.

An editorial from Mannino and Townsend (21) questions whether a single set of “racial-composite” prediction equations is the best path toward precision medicine and encourages the study of the applications of GLI Global before endorsing widespread adoption. Our paper adds data to support the use of the GLI Global reference equations.

Notably, the severity of airflow obstruction needs to be taken into consideration when evaluating surgical options (e.g., lung volume reduction or lung transplant) for COPD. The International Society for Heart and Lung Transplantation (22) recommends referral to a lung transplant center for a ppFEV₁ <25. Using a race-neutral reference equation would reduce the ppFEV₁ in Black patients, triggering an earlier referral than would occur with the use of race-specific equations (7). However, as mentioned in the ATS 2023 statement, it is unknown whether the potential changes in lung transplant referral patterns will affect survival positively or negatively in Black or White patients. Brems and coworkers (23) explored the impact of race-specific spirometry interpretation on Lung Allocation Score, the primary determinant of lung transplant priority, with a higher score indicating the patient is more likely to benefit, therefore increasing the chances of receiving an organ. In a cohort of 8,982 patients (90.3% and 9.7% of White and Black race, respectively), shifting from a race-specific (GLI) to a race-neutral (GLI other) approach, the ppFVC increased 4.4% in White patients and decreased 3.8% in Black patients. This shift determined a decrease in Lung Allocation Score of 0.6 points in White patients and an increase of 0.6 points in Black patients. The investigators concluded that a race-specific approach has the potential to adversely affect the care of Black patients with advanced respiratory disease and may have contributed to racially biased allocation of lung transplants. Our results suggest that using the zGLI Global severe class would refer a greater number of patients for lung transplant evaluation. These findings also confirm that the use of a race-neutral reference equation would increase the number and proportion of eligible African American subjects and would not reduce the overall count of White individuals.

Our study has several strengths. First, the results were obtained in a large sample size in which detailed phenotyping and long-term follow-up were available. Moreover, similar proportions of male and female participants and a substantial number of African American participants were enrolled. Second, our study explored the application of the 2021 ERS/ATS recommendations and implemented the race-neutral reference equations, as suggested in the 2023 document. Furthermore, we separately analyzed the contributions of using z-scores and LLNs (step 1) and the shift from race-specific to race-neutral reference equations (step 2). To our knowledge, no study has undertaken this combined approach. Last, for all the intraclass comparisons in this study, the reference group included only subjects with normal spirometry and did not include subjects defined as having PRISm. This choice was supported by the evidence that, even though neither PRISm nor GOLD stage 0 (and, equivalently, zGLI pre-COPD) are currently defined as COPD, these two groups are different in symptoms and progression, justifying a distinction (24, 25).

Our study is limited in that COPDGene included only subjects who self-identified as non-Hispanic White or African American and were enrolled in the United States. Using the GLI Global reference equations could potentially encourage the enrollment of multiracial participants in clinical studies. Our findings would also need to be replicated in other studies that include other racial groups in the United States and worldwide.

We acknowledge that pulmonary function test results obtained at a single time point are not immune from the uncertainty of measurement and interpretation, particularly around the thresholds. This limitation, however, is consistent with the usual clinical scenario of everyday practice, in which the clinician is required to classify patients into discrete categories based on a single set of spirometry data. Nevertheless, we acknowledge that future studies should consider longitudinal follow-up.

In conclusion, our findings support the hypothesis that a severity classification based on the 2021 ERS/ATS standards and GLI Global reference equations could discriminate survival, imaging characteristics, and exacerbation rates better than the GOLD classification. Further studies are needed to address the effects of social determinants of health on COPD morbidity and mortality.