To the Editor:
Racial and ethnic variation in normal lung function can influence the interpretation of spirometric measurements. The Global Lung Function Initiative (GLI) derived multiethnic reference equations in large, representative populations of healthy nonsmokers to account for differences in lung function among racial and ethnic groups (1). Genetic ancestry underlies part of these differences (2, 3), and the GLI equations make no attempt to account for variation in genetic ancestry among admixed racial and ethnic groups (4). Herein we demonstrate how individuals’ measures of genetic ancestry can be used for the more precise application of the GLI equations.
Methods
Spirometric predictions and genetic ancestry proportions were calculated using data collected in two parallel case-control studies of asthma (2006–2014): the GALA II (Genes-Environments and Admixture in Latino Americans) study and SAGE (Study of African Americans, Asthma, Genes, and Environments) (5). Control subjects were healthy with no history of lung disease. Cases had physician-diagnosed asthma. Subjects were 8–21 years old, and their parents and grandparents self-identified as Hispanic or Latino (Mexican American, Puerto Rican, or other Latino) or African American. Spirometry was performed per American Thoracic Society recommendations (6). As previously described (3), ADMIXTURE software was used for inferring genetic ancestry based on genotypes at a large number of SNP markers. FEV1 and FVC predictions, percent predicted, and z-scores were calculated using the GLI equations derived in White and African American individuals, and the composite equation derived for multiracial or unrepresented populations. Complete spirometry and genome-wide genetic data were available for 1,049 control and 3,132 case individuals.
In a random sample of controls (training dataset, n = 699), each individual was categorized according to whether they were best fit (smallest z-score absolute value) with the White or composite equation, and, separately, these individuals were categorized according to whether they were best fit by the composite or African American equation. The association between best fit equation and African ancestry was estimated using logistic regression among control subjects in the training dataset, unadjusted and adjusted for race and ethnicity. Unadjusted logistic regressions were used to determine the African ancestry cutoff point probabilities at which an individual was more likely to be best fit by the White or composite equation (model 1) and cutoff points at which an individual was more likely to be best fit by the composite or African American equation (model 2) among controls in the training dataset. Resulting African ancestry cutoff points formed the basis for genetic ancestry–informed equation selection, which was evaluated in the remaining control subjects (test dataset, n = 350) and compared with race/ethnicity–based equation selection and a “one-size-fits-all” approach using the composite equation for all subjects. Using the test dataset, the equation selection method with the smallest mean z-score absolute value was deemed best fit with “sufficient fit” if that value was less than 0.5 using two one-sided t tests for equivalence (7). Method performance was also evaluated via classification accuracy (correct prediction being the best fit equation) and root-mean-square error.
Among asthma cases, ancestry-informed equation selection was compared with race/ethnicity– and composite-based approaches for classifying severity of spirometric abnormality based on FEV1 percent predicted (8).
Results
In the training dataset, for every 10% increase in African ancestry, the odds of an individual being best fit by the GLI composite equation compared with the White equation was 1.27 (95% confidence interval [CI], 1.20–1.34) for FEV1 and 1.28 (95% CI, 1.22–1.36) for FVC. For every 10% increase in African ancestry, the odds of an individual being best fit by the GLI African American equation compared with the composite equation was 1.27 (95% CI, 1.20–1.34) for FEV1 and 1.24 (95% CI, 1.18–1.31) for FVC. Associations were robust to adjustment for race and ethnicity (data not shown in tables).
An individual was most likely to be best fit by the White equation if African ancestry was less than 32.5% for FEV1 and less than 32.8% for FVC. An individual was most likely to be best fitby the composite equation if African ancestry was between 32.5 and 78.2% for FEV1 and between 32.8 and 77.2% for FVC. An individual was most likely to be best fit by the African American equation if African ancestry was more than 78.2% for FEV1 and more than 77.2% for FVC (data not shown in tables).
Genetic ancestry–informed equation selection in the test dataset resulted in superior fit (FEV1 and FVC mean z-scores of 0.04 [95% CI, −0.07 to 0.14] and −0.03 [95% CI, −0.08 to 0.14] with root-mean-square error of 0.35 for FEV1 and 0.43 for FVC) and greater classification accuracy (0.58 [95% CI, 0.52 to 0.63] and 0.53 [95% CI, 0.48 to 0.59] for FEV1 and FVC) than when using the race/ethnicity–based approach to equation selection or using the composite equation for all (Table 1).
Table 1.
Characteristics of Healthy Control Participants in GALA II and SAGE and Global Lung Function Initiative Equation Fit Based on Genetic Ancestry–informed, Race/Ethnicity–based, and “One-Size-Fits-All” Approaches to Equation Selection: 2006–2014
| Training Dataset (n = 699) | Test Dataset (n = 350) | Total (N = 1,049) | |
|---|---|---|---|
| Boys, n (%) | 289 (41.3) | 146 (41.7) | 435 (41.5) |
| Age, yr, mean (SD) | 15.4 (3.6) | 15.2 (3.6) | 15.3 (3.6) |
| Height, cm, mean (SD) | 159.6 (12.4) | 158.4 (12.8) | 159.2 (12.6) |
| Weight, kg, mean (SD) | 62.7 (21.1) | 62 (21.7) | 62.4 (21.3) |
| African ancestry, %, mean (SD) | 31.2 (31.5) | 31.0 (31.5) | 31.1 (31.5) |
| African American, n (%) | 187 (26.8) | 97 (27.7) | 284 (27.1) |
| Hispanic or Latino, n (%) | 512 (73.2) | 253 (72.3) | 765 (72.9) |
| FEV1, L, mean (SD) | 3.08 (0.88) | 3.00 (0.83) | 3.06 (0.86) |
| FVC, L, mean (SD) | 3.52 (1.06) | 3.43 (0.97) | 3.49 (1.03) |
| FEV1 fit | |||
| Genetic ancestry informed* | |||
| z-Score, mean (95% CI) | —† | 0.04 (−0.07 to 0.14)‡ | —† |
| RMSE | —† | 0.35 | —† |
| Classification accuracy (95% CI)§ | —† | 0.58 (0.52 to 0.63) | —† |
| Race/ethnicity based‖ | |||
| z-Score, mean (95% CI) | 0.41 (0.32 to 0.50) | 0.34 (0.23 to 0.45)‡ | 0.39 (0.32 to 0.46) |
| RMSE | 0.45 | 0.38 | 0.43 |
| Classification accuracy (95% CI)§ | 0.42 (0.39 to 0.46) | 0.45 (0.39 to 0.50) | 0.43 (0.40 to 0.46) |
| Composite based | |||
| z-Score, mean (95% CI) | 0.43 (0.34 to 0.53) | 0.35 (0.23 to 0.47)‡ | 0.41 (0.33 to 0.48) |
| RMSE | 0.48 | 0.41 | 0.45 |
| Classification accuracy (95% CI)§ | 0.22 (0.19 to 0.25) | 0.21 (0.17 to 0.26) | 0.22 (0.19 to 0.24) |
| FVC fit | |||
| Genetic ancestry informed* | |||
| z-Score, mean (95% CI) | —† | 0.03 (−0.08 to 0.14)‡ | —† |
| RMSE | —† | 0.43 | —† |
| Classification accuracy (95% CI)§ | —† | 0.53 (0.48 to 0.59) | —† |
| Race/ethnicity based‖ | |||
| z-Score, mean (95% CI) | 0.39 (0.30 to 0.48) | 0.37 (0.26 to 0.49)‡ | 0.38 (0.31 to 0.45) |
| RMSE | 0.52 | 0.45 | 0.50 |
| Classification accuracy (95% CI)§ | 0.43 (0.39 to 0.47) | 0.44 (0.39 to 0.49) | 0.43 (0.40 to 0.46) |
| Composite based¶ | |||
| z-Score, mean (95% CI) | 0.47 (0.37 to 0.57) | 0.43 (0.30 to 0.55) | 0.46 (0.38 to 0.53) |
| RMSE | 0.56 | 0.47 | 0.53 |
| Classification accuracy (95% CI)§ | 0.20 (0.17 to 0.23) | 0.23 (0.19 to 0.28) | 0.21 (0.19 to 0.24) |
Definition of abbreviations: CI = confidence interval; GALA II = Genes-Environments and Admixture in Latino Americans; RMSE = root-mean-square error; SAGE = Study of African Americans, Asthma, Genes, and Environments.
Genetic ancestry–informed selection of Global Lung Function Initiative equations involved use of the White-derived equation if an individual’s African ancestry was less than 32.5% for FEV1 and 32.8% for FVC, use of the African American–derived equation if African ancestry was more than 78.2% and 77.2% for FEV1 and FVC, respectively, and use of the composite equation for individuals with African ancestry between these cutoff points. African ancestry cutoff points were derived in the training dataset and assessed in the test dataset.
Performance metrics were not calculated in the training and total datasets because the models for a genetic ancestry–informed approach to equation selection were derived in the training dataset.
Sufficient fit with absolute value of mean z-score less than 0.5 using two one-sided t tests for equivalence.
Classification accuracy is the ratio of predictions that are the best fit equation (correct predictions) to the total number of predictions.
Race/ethnicity–based selection involved use of the African American–derived equation for African American individuals, the White-derived equation for Mexican American individuals, and the composite equation for Puerto Rican and other Latino individuals.
Composite-based selection involved a “one-size-fits-all” approach whereby the composite equation was used for all subjects regardless of race/ethnicity or genetic ancestry.
A total of 162 (5.2%) cases were classified as having a moderate-to-very severe spirometric abnormality based on FEV1 percent predicted using ancestry-informed equation selection, compared with 115 (3.7%) and 125 (4.0%) using race/ethnicity–based or composite-based equation selection, respectively (Table 2).
Table 2.
Characteristics of Asthma Case Participants in GALA II and SAGE and Classification of Disease Severity Using Genetic Ancestry–informed, Race/Ethnicity–based, and “One-Size-Fits-All” Approaches to Equation Selection: 2006–2014
| African American (n = 931) |
Hispanic or Latino (n = 2,201) |
Total (N = 3,132) | |
|---|---|---|---|
| Boys, n (%) | 470 (50.5) | 1,194 (54.2) | 1,664 (53.1) |
| Age, yr, mean (SD) | 14.2 (3.7) | 12.8 (3.3) | 13.2 (3.5) |
| Height, cm, mean (SD) | 158 (14.5) | 151 (14) | 153.1 (14.5) |
| Weight, kg, mean (SD) | 63.6 (24.2) | 54.9 (21.7) | 57.5 (22.8) |
| African ancestry, %, mean (SD) | 78.5 (11.9) | 16.3 (13.4) | 34.8 (31.2) |
| FEV1, L, mean (SD) | 2.57 (0.78) | 2.47 (0.83) | 2.5 (0.82) |
| FVC, L, mean (SD) | 3.11 (0.97) | 2.92 (0.99) | 2.98 (0.99) |
| FEV1 % pred, mean (SD) | |||
| Genetic ancestry informed | 0.96 (0.14) | 0.94 (0.16) | 0.95 (0.15) |
| Race/ethnicity based | 0.99 (0.14) | 0.98 (0.16) | 0.99 (0.16) |
| Composite based | 0.91 (0.13) | 1.00 (0.17) | 0.98 (0.16) |
| Mild (FEV1 %pred ≥ 70%), n (%) | |||
| Genetic ancestry informed | 903 (97.0) | 2,067 (93.9) | 2,970 (94.8) |
| Race/ethnicity based | 905 (97.2) | 2,111 (95.9) | 3,016 (96.3) |
| Composite based | 892 (95.8) | 2,115 (96.1) | 3,007 (96.0) |
| Moderate (60% ≤ FEV1 %pred < 70%), n (%) | |||
| Genetic ancestry informed | 21 (2.3) | 97 (4.4) | 118 (3.8) |
| Race/ethnicity based | 20 (2.1) | 65 (3.0) | 85 (2.7) |
| Composite based | 21 (2.3) | 61 (2.8) | 82 (2.6) |
| Moderately severe (50% ≤ FEV1 %pred < 60%), n (%) | |||
| Genetic ancestry informed | 5 (0.5) | 26 (1.2) | 31 (1.0) |
| Race/ethnicity based | 4 (0.4) | 17 (0.8) | 21 (0.7) |
| Composite based | 14 (1.5) | 19 (0.9) | 33 (1.1) |
| Severe (35% ≤ FEV1 %pred < 50%), n (%) | |||
| Genetic ancestry informed | 2 (0.2) | 10 (0.5) | 12 (0.4) |
| Race/ethnicity based | 2 (0.2) | 7 (0.3) | 9 (0.3) |
| Composite based | 4 (0.4) | 5 (0.2) | 9 (0.3) |
| Very severe (FEV1 %pred <, 35%), n (%) | |||
| Genetic ancestry informed | 0 (0.0) | 1 (0.0) | 1 (0.0) |
| Race/ethnicity based | 0 (0.0) | 1 (0.0) | 1 (0.0) |
| Composite based | 0 (0.0) | 1 (0.0) | 1 (0.0) |
Definition of abbreviations: %pred = percent predicted; GALA II = Genes-Environments and Admixture in Latino Americans; SAGE = Study of African Americans, Asthma, Genes, and Environments.
FEV1 percent predicted was calculated using ancestry-informed selection of Global Lung Function Initiative equations, race/ethnicity–based selection, and a “one-size-fits-all” approach whereby the composite equation was used for all subjects regardless of race/ethnicity of genetic ancestry.
Discussion
In a population of African American and Hispanic or Latino children, applying genetic ancestry to the selection of individuals’ GLI reference equation provided advantages over equation selection based on race/ethnicity or a one-size-fits-all approach wherein the composite equation, derived for multiracial or unrepresented populations, was used for all participants. Ancestry-informed equation selection resulted in superior fit among healthy children and a greater proportion of children with asthma being classified as having moderate-to-very severe spirometric abnormalities. These findings have important implications given that misestimated lung function can result in errors and delays in disease detection and medical management, misclassification of disease severity, denial of disability claims, exclusion from clinical trials, and ineligibility for lifesaving treatments such as transplants and other surgeries.
Consistent with a previous study, we observed that normal lung function among healthy African American and Mexican American individuals are, respectively, lower than and similar to lung function among healthy White individuals (1). Our findings also confirm previous observations that race/ethnicity–based equations can misestimate lung function after accounting for intrapopulation variation in African ancestry (2, 4). Although the use of a one-size-fits-all approach improved sensitivity for detecting moderate-to-very severe spirometric abnormalities among African American children with asthma, our overall findings support prior claims that eliminating the use of race/ethnicity in lung function prediction could exacerbate racial/ethnic inequities in health outcomes (9). This is of critical concern given that Black or African American and Hispanic or Latino children collectively make up approximately 40% of children in the United States (10). Additional studies are, therefore, needed to further assess how the use of genetic ancestry–informed, race/ethnicity–based, and one-size-fits-all spirometry reference equations influence health outcomes among diverse populations.
We have previously demonstrated that African genetic ancestry is inversely related to lung function, even after accounting for social determinants of health such as early life exposures, air pollution, and socioeconomic status (2, 3). Herein, we show that using individuals’ African genetic ancestry for the selection of race/ethnicity–based spirometry reference equations improves the precision of lung function prediction. However, our analysis was performed exclusively in children, thus requiring further studies in adults to replicate our findings. In addition, our findings must be replicated in other populations of Black or African American and Hispanic or Latino individuals, as well as in other racial/ethnic groups. Finally, further studies are needed to examine the implications and practicality of incorporating genetic ancestry in pulmonary function testing. Despite these limitations, recent scientific advances allow for estimates of genetic ancestry to be easily and inexpensively measured. Accordingly, technological limitations are no longer a justifiable reason for not using genomics in the pulmonary function laboratory. Promoting race-conscious medicine demands that we bring precision medicine to disadvantaged populations without reverting to a less precise one-size-fits-all approach, which can bring greater disadvantage than any proposed benefit.
Acknowledgment:
The authors thank the participants and their families for their contribution, as well as the healthcare professionals and clinics for their support and participation in the Genes-Environments and Admixture in Latino Americans Study and the Study of African Americans, Asthma, Genes and Environments.
Supported in part by the Sandler Family Foundation; the American Asthma Foundation; the RWJF Amos Medical Faculty Development Program; Harry Wm. and Diana V. Hind Distinguished Professor in Pharmaceutical Sciences II; NIH, NHLBI (R01HL117004, 1X01HL134589, R01HL128439, R01HL135156, R01HL141992, and R01HL155024); the National Institute of Environmental Health Sciences (R01ES015794 and R21ES24844); the National Institute on Minority Health and Health Disparities (R01MD010443 and R56MD013312); the National Human Genome Research Institute (U01HG009080); the National Institute of General Medical Sciences (T32GM007546); the National Center for Advancing Translational Sciences, NIH (through UCSF-CTSI grant number TL1 TR001871); and the UCSF Pediatric Allergy, Immunology, and Bone Marrow Transplantation Division. The content of this publication is solely the responsibility of the authors and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.
Footnotes
Author disclosures are available with the text of this letter at www.atsjournals.org.
Contributor Information
Jonathan Witonsky, University of California San Francisco San Francisco, California.
Jennifer R. Elhawary, University of California San Francisco San Francisco, California
Celeste Eng, University of California San Francisco San Francisco, California
José R. Rodríguez-Santana, Centro de Neumología Pediátrica San Juan, Puerto Rico
Luisa N. Borrell, City University of New York New York, New York.
Esteban G. Burchard, University of California San Francisco San Francisco, California.
References
- 1.Quanjer PH, Stanojevic S, Cole TJ, Baur X, Hall GL, Culver BH, et al. ; ERS Global Lung Function Initiative. Multi-ethnic reference values for spirometry for the 3–95-yr age range: the global lung function 2012 equations. Eur Respir J 2012;40:1324–1343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kumar R, Seibold MA, Aldrich MC, Williams LK, Reiner AP, Colangelo L, et al. Genetic ancestry in lung-function predictions. N Engl J Med 2010;363:321–330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Pino-Yanes M, Thakur N, Gignoux CR, Galanter JM, Roth LA, Eng C, et al. Genetic ancestry influences asthma susceptibility and lung function among Latinos. J Allergy Clin Immunol 2015;135:228–235. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Witonsky J, Elhawary JR, Eng C, Rodríguez-Santana JR, Borrell LN, Burchard EG. Race- and ethnicity-based spirometry reference equations: are they accurate for genetically admixed children? Chest [online ahead of print] 13 Jan 2022; DOI: 10.1016/j.chest.2021.12.664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Nishimura KK, Galanter JM, Roth LA, Oh SS, Thakur N, Nguyen EA, et al. Early-life air pollution and asthma risk in minority children. The GALA II and SAGE II studies. Am J Respir Crit Care Med 2013;188:309–318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, Coates A, et al. ; ATS/ERS Task Force. Standardisation of spirometry. Eur Respir J 2005;26:319–338. [DOI] [PubMed] [Google Scholar]
- 7.Hall GL, Thompson BR, Stanojevic S, Abramson MJ, Beasley R, Coates A, et al. The Global Lung Initiative 2012 reference values reflect contemporary Australasian spirometry. Respirology 2012;17:1150–1151. [DOI] [PubMed] [Google Scholar]
- 8.Pellegrino R, Viegi G, Brusasco V, Crapo RO, Burgos F, Casaburi R, et al. Interpretative strategies for lung function tests. Eur Respir J 2005;26:948–968. [DOI] [PubMed] [Google Scholar]
- 9.Borrell LN, Elhawary JR, Fuentes-Afflick E, Witonsky J, Bhakta N, Wu AHB, et al. Race and genetic ancestry in medicine - a time for reckoning with racism. N Engl J Med 2021;384:474–480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Kids Count Data Center. Child population by race in the United States. The Annie E. Casey Foundation; 2021 [updated 2021. Sep; accessed 2021 Nov 4]. Available from: https://datacenter.kidscount.org/data/tables/103-child-population-by-race?loc=1&loct=1#detailed/1/any/false/1729,37,871,870,573,869,36,868,867,133/68,69,67,12,70,66,71,72/423,424. [Google Scholar]