FEV1Q: a Race-Neutral Approach to Assessing Lung Function

Abstract

Rationale:

FEV1Q is a simple approach to spirometry interpretation that compares measured lung function to a lower boundary. This study evaluated how well FEV1Q predicts survival compared to current interpretation methods and whether race impacts FEV1Q.

Methods:

White and Black adults with complete spirometry and mortality data from the National Health and Nutrition Examination Survey (NHANES) III and United Network for Organ Sharing (UNOS) database for lung transplant referrals were included. FEV1Q was calculated as forced expiratory volume in one second (FEV1) divided by 0·4L for females or 0·5L for males. Cumulative distributions of FEV1 were compared across races. Cox proportional hazard models tested mortality risk from FEV1Q adjusting for age, sex, height, smoking, income, and among UNOS individuals, referral diagnosis. Harrell’s C-statistics were compared between absolute FEV1, FEV1Q, FEV1/height², FEV1 z-scores and FEV1 % predicted. Analyses were stratified by race.

Results:

Among 7182 individuals from NHANES and 7149 from UNOS, 1907 (27%) and 991 (14%) were Black respectively. The lower boundary FEV1 values did not differ between Black and White individuals in either population (FEV1 1^st percentile difference: ≤0·01L; p-value>0·05). Decreasing FEV1Q was associated with increasing hazard ratio (HR) for mortality (NHANES HR 1·33, 95% CI 1·28–1·39; UNOS HR 1·18, 95% CI 1·12–1·23). The associations were not confounded nor modified by race. Discriminative power was highest for FEV1Q compared to alternative FEV1 approaches in both Black and White individuals.

Conclusions:

FEV1Q is an intuitive and simple race-neutral approach to interpreting FEV1 that predicts survival better than current alternative methods.

Introduction

Interpreting spirometry remains the cornerstone of respiratory health assessment. An ideal interpretive strategy should be 1) simple to understand, 2) easy to implement, 3) avoid inequity across patient groups and 4) accurately reflect key clinical outcomes such as survival and functional status. Current interpretive standards recommend comparing forced expiratory volume in one second (FEV1) to a predicted value to calculate a z-score¹. Predicted values are estimated using reference equations developed from healthy populations and incorporate age, sex, height, and until recently, race^2,3. This approach has been challenged as potentially being inequitable with respect to race^4–7.

Another approach, the FEV1 quotient (FEV1Q), posits that individuals have a lower boundary of lung function that approximates a lower limit needed to survive which is 0·4L for women and 0·5L for men⁸. FEV1Q anchors interpretation to that boundary rather than a predicted ‘normal’ value and is calculated as FEV1 divided by the sex specific minimum, effectively describing how much lung function an individual has left before death is highly likely⁸. FEV1Q is easy to understand and implement, and has been associated with mortality, chronic obstructive pulmonary disease (COPD) exacerbations, and functional status^8–13. The recent European Respiratory Society/American Thoracic Society interpretation standards recommended further evaluation of FEV1Q as an interpretation strategy¹, with some key knowledge gaps that include understanding the utility of FEV1Q in diseases aside from COPD and in more diverse populations.

Recent evidence has suggested that applying race specific reference equations may falsely underestimate disease severity in Black and Asian populations, prompting the call for a single interpretation strategy across racial and ethnic groups^3,14–17. At the same time, epidemiologic data has demonstrated differences in average lung function across racial groups^2,18–24, and the societal implications of applying a single race-neutral reference equation to all individuals^25–27 have led to disagreement over which approach is the most appropriate. The focus of this debate relies heavily on defining an individual’s predicted value and selecting appropriate healthy reference populations. A method that anchors to a lower boundary, such as FEV1Q, has the potential to avoid this debate. However, it is currently unknown whether there are differences in the observed lower boundaries for FEV1 by race.

The goal of this study was to expand the evaluation of FEV1Q to more diverse populations representing both healthy and diseased individuals. Further, this study examined whether the lower boundary of FEV1 varies by race, and whether race alters the association between FEV1Q and mortality.

Methods

Study Populations

The NHANES III (National Health and Nutrition Examination Survey III) population was used to study a general population and the United Network for Organ Sharing (UNOS) population to study people with advanced lung disease. The NHANES III collected data between 1988 and 1994, with mortality data through 2015^28,29 and we restricted analyses to those self-identified as White or Black who were over age 40, as previous work demonstrated a stable first percentile FEV1 value over the age of 40⁸. The UNOS population consisted of individuals with advanced lung disease who had been listed for lung transplant after May 2005, with follow up and mortality data through September 2022. Adults over age 18 years who self-identified as White or Black with FEV1 data obtained immediately prior to listing were included in analyses.

Spirometry Interpretive Strategies

The primary methods for interpreting FEV1 included (1) absolute FEV1 values in litres, (2) FEV1Q, (3) FEV1 z-scores, and (4) FEV1 % predicted values. FEV1Q values were defined as an individual’s absolute FEV1 divided by 0·4L for females or 0·5L for males⁸. All FEV1 z-scores and FEV1 % predicted values were calculated using Global Lung Initiative (GLI) race-specific reference equations and with GLI Global reference equations for everyone as a race-neutral approach^2,3.

Survival Data

NHANES III has linked survival data from the date of spirometry to date of death or December 31^st, 2015. Cause of death is provided as one of nine categories: heart disease, cancer, chronic lower respiratory disease, unintentional injuries, cerebrovascular diseases, Alzheimer’s disease, diabetes, pneumonia and influenza, and kidney disease. Analyses of respiratory-specific cause of death included those coded as chronic lower respiratory disease or pneumonia and influenza.

For UNOS mortality analyses, individuals who were transplanted were excluded to avoid this competing outcome. Follow up time was calculated as time from index FEV1 measurement to date of death, date of removal from the waitlist, or September 30th, 2022.

Statistical Analysis

Statistical analyses were conducted using STATA 15·1. Clinical characteristics were summarised using means and standard deviations, or percentages as appropriate. The NHANES data was analysed using survey weights, sampling units, and strata^28,29. First percentile values of FEV1 were visualised using the UNOS data and estimates of confidence intervals around the first percentile value were obtained using a binomial method³⁰. Comparisons of relative percentiles across races in both the NHANES and UNOS data were visualised using cumulative distribution plots. To test statistical differences in FEV1 values at each percentile, generalised gamma models were fitted to each cumulative distribution curve^31,32.

Kaplan-Meier curves were used to visualise differences in FEV1 interpretive strategy on the associations between lung function quartiles and all-cause mortality. Cumulative incidence of respiratory-specific death over time by FEV1Q was analysed with FEV1Q dichotomised at its median value. Non-respiratory causes of death were treated as a competing outcome to respiratory-specific mortality in this analysis. Harrell’s C-statistics, calculated using a jackknife approach^33–35, were used to compare the discriminative performance of FEV1Q against alternative FEV1 metrics. Analyses were stratified by race.

Multivariable Cox proportional hazard models were generated to test associations between FEV1Q and mortality. Proportional hazards assumptions were verified in both study populations. For NHANES models, covariates were age, sex, standing height, smoking status, smoking pack-years, and income to poverty ratio. For UNOS models, covariates were age, sex, standing height, smoking status, and lung transplant referral diagnosis group. A base model was evaluated followed by a model including a race term and an interaction term between race and FEV1Q. Age at death in NHANES or years survival on the waitlist in UNOS were estimated by race across FEV1Q bins. An additional sensitivity analysis of NHANES models was performed including adults over 18 years.

Results

Clinical Characteristics

There were 7,182 individuals from NHANES and 33,174 from UNOS included in analyses of distributions of FEV1. For the mortality analyses there were 7,182 individuals from NHANES and 7,149 from UNOS who had not been transplanted (Supplemental Figure 1). In both populations, White individuals were most prevalent followed by Black, with a survey weight of 85% White in NHANES, and 89% White in UNOS. Fifty-five percent of NHANES and 29% of UNOS populations died during follow-up with mean follow-up times of 18·0 and 1·1 years, respectively.

After applying sampling weights for NHANES, Black individuals tended to be younger, more frequently female, and had a lower income to poverty ratio than White individuals (Table 1). There were no observed differences by race in standing height, and Black individuals were more commonly never or current smokers with lower pack-years than White individuals. In the UNOS population there were no differences in height, but White individuals tended to be older, more likely to be working for income at the time of listing for lung transplant and less likely to be never smokers than Black individuals (Table 1). In UNOS there were racial differences in diagnosis with greater prevalence of Cystic Fibrosis in the White, and greater prevalence of Pulmonary Vascular Disease among Black individuals.

Table 1:

Clinical characteristics by race in NHANES and UNOS study populations.

	NHANES		UNOS
Characteristic	White N=5275	Black N=1907	White N=29,481	Black N=3,693
Survey Weight Proportion, %	85	11	--	--
Age (yrs)	57 ± 0·4	55 ± 0·4	57 ± 13	53 ± 11
Female, %	52	56	41	57
Standing Height (cm)	168 ± 0·2	168 ± 0·2	170 ± 10	169 ± 10
Smoking status, %
Never	39	42	39	48
Former	38	24	61	52
Current	23	34	0	0
Smoking Pack-years	21 ± 0·5	14 ± 0·5	--	--
Household Income to Poverty Ratio	3.6 ± 0·09	2.3 ± 0·07	--	--
Working for Income at Listing, N (%)*	--	--	3,586 (16)	332 (13)
Respiratory Diagnosis, N (%)
Obstructive Lung	--	--	8608 (29)	956 (26)
Pulmonary Vascular	--	--	1254 (4)	294 (8)
Cystic Fibrosis	--	--	2868 (10)	66 (2)
Restrictive Lung	--	--	16751 (57)	2377 (64)

For NHANES the N values are raw numbers, but all summary statistics are after applying survey weights. Smoking pack-years not available for UNOS data. Working for income and lung transplant referral diagnosis data are only available for the UNOS dataset. Statistics are mean ± standard deviation, or percentage as appropriate.

^*Working for income data available in N=25,487

FEV1 Demonstrates a Stable Lower Boundary Compatible with Survival Across Races

Using the UNOS population we verified a stable lower boundary for FEV1 (Supplemental Figure 2), which was lower in females than in males. As the UNOS population have very severe lung disease, there were no differences by age group, as might be expected among younger adults in a healthy population. The lower bounds of FEV1 did not vary between White and Black individuals but were slightly lower than previously found⁸. Further, after stratifying by sex, the lower boundary did not vary by height (Supplemental Figure 3).

Examining the cumulative distributions of FEV1 in the mostly healthy NHANES (Figure 1A) and the severely diseased in UNOS (Figure 1B) there was no difference in first percentile values between Black and White individuals. In NHANES the 95th percentile for FEV1 was 3·8L in Blacks and 4·3L in Whites with divergence in distributions occurring above the 5th percentile (Supplemental Table 1). In UNOS the 95th percentile for FEV1 was also lower in Blacks with divergence of distributions above 1L. Sensitivity analysis including adults over 18 in NHANES confirmed that cumulative distributions of FEV1 by race were very similar to those over age 40 (Supplemental Figure 4).

Figure 1.

Cumulative distributions of FEV1 among White and Black individuals from the NHANES and UNOS study populations

FEV1 cumulative distributions for White individuals are represented by solid lines, and Black individuals are represented by dashed lines. The NHANES study population is in panel A and the UNOS study population in panel B.

FEV1Q is a strong predictor of mortality and is neither confounded by nor modified by race

The healthier population of NHANES was used to compare associations with mortality and the various FEV1 metrics. Individuals with the same z-score were observed to have a wide range of absolute FEV1Q values (Supplemental Figure 5) with FEV1 z-scores between 0 and +1 being compatible with FEV1Q values ranged from 3·8 to 11·5. Figure 2 shows Kaplan-Meier curves of FEV1Q quartiles (panel A) compared to FEV1 z-score quartiles (panel B). FEV1Q demonstrated separation of mortality risk between all four quartiles while FEV1 z-scores (panel B) and % predicted values (Supplemental Figure 6) demonstrated mortality risk separation only for the lowest quartile. Race did not affect the association between FEV1Q and all-cause mortality (Supplemental Figure 7) or respiratory mortality (Supplemental Figure 8).

Figure 2.

Kaplan-Meier curves of survival comparing FEV1Q and FEV1 z-score quartiles

Survival curves from the NHANES study population using FEV1Q quartiles (panel A) and FEV1 z-score quartiles (panel B).

Harrell’s C-statistics were calculated to compare the discriminative performance of FEV1Q with other interpretive strategies for spirometry, including absolute FEV1, FEV1/height², both race-specific and race-neutral z-scores and % predicted values. FEV1Q had the best performance with a statistically higher C-statistic of 0·7 (Figure 3), outperforming absolute FEV1 with a C-statistic of 0·65. Applying a Bonferroni correction to set significance at a p<0.008, FEV1Q continued to outperform all other strategies except FEV1/height² (p=0.009). When stratified by race there was no difference in C-statistics, with FEV1Q surpassing all other FEV1 measures (p<0·05).

Figure 3.

Comparison of Harrell’s C-statistics across FEV1 interpretive strategies.

Harrell’s C-statistics for univariate Cox regression models in the NHANES population for each method of FEV1 interpretation.

The relationship between FEV1Q and mortality in the UNOS population was found to be highly associated with the referral diagnosis. In the NHANES population, univariate analysis demonstrated that higher FEV1Q was associated with lower risk of death while in the UNOS population univariate analysis for FEV1Q demonstrated the opposite association. However, when adjusted for lung transplant referral diagnosis, the association was similar to that of the NHANES population (Supplemental Table 2).

Table 2 demonstrates two multivariable models for both populations; the first excluding race as a covariate and the second including a race term and an interaction term to examine whether race modified the FEV1Q and mortality association. Neither the race term nor the interaction term was significant in the models for either population. Further, the magnitude of association for FEV1Q did not change across the models that included or excluded race terms, demonstrating that race neither confounded nor modified the association between FEV1Q and mortality across both a population with severe lung disease and a healthier general population. A sensitivity analysis including adults in NHANES 18–40 years, demonstrated similar results (Supplemental Table 3). Thus, we observed in both NHANES and UNOS populations, adjusting for key confounders, predicted survival was worse with lower FEV1Q values, and this did not significantly differ by race (Supplemental Figure 9).

Table 2:

Hazard ratios for all-cause mortality from Cox regression examining the association between FEV1Q and mortality in the NHANES and UNOS datasets.

NHANES N=6380					UNOS N=7025
	Model 1		Model 2			Model 1		Model 2
	HR (95% CI)	p-value	HR (95% CI)	p-value		HR (95% CI)	p-value	HR (95% CI)	p-value
FEV1Q	0·76 (0·73–0·79)	<0·001	0·75 (0·72–0·78)	<0·001	FEV1Q	0·85 (0·81–0·89)	<0·001	0·85 (0·81–0·89)	<0·001
Black vs White Race	-	-	0·64 (0·38–1·08)	0·097	Black vs White Race	-	-	1·00 (0·72–1·37)	0·993
FEV1Q * Black Race	-	-	1·08 (0·98–1·2)	0·13	FEV1Q * Black Race	-	-	0·93 (0·83–1·04)	0·186
Age (yrs)	1·09 (1·09–1·09)	<0·001	1·09 (1·08–1·10)	<0·001	Age (yrs)	1·01 (1·00–1·01)	<0·001	1·01 (1·00–1·01)	<0·001
Female Sex	0·73 (0·64–0·83)	<0·001	0·74 (0·65–0·84)	<0·001	Female Sex	0·70 (0·62–0·79)	<0·001	0·72 (0·63–0·81)	<0·001
Height (cm)	1·01 (1·01–1·02)	0·001	1·01 (1·01–1·02)	0·001	Height (cm)	1·00 (0·99–1·01)	0·218	1·00 (0·99–1·01)	0·121
Former vs Never	1·15 (1·02–1·30)	0·022	1·15 (1·02–1·30)	0·022	Ever vs Never	0·98 (0·89–1·09)	0·764	0·98 (0·89–1·08)	0·664
Current vs Never	1·78 (1·53–2·08)	<0·001	1·79 (1·53–2·09)	<0·001	PVD	3·61 (2·92–4·47)	<0·001	3·69 (2·98–4·58)	<0·001
Pack-years	1·00 (1·00–1·01)	<0·001	1·00 (1·00–1·01)	<0·001	CF	3·19 (2·58–3·95)	<0·001	3·07 (2·47–3·80)	<0·001
Income	0·92 (0·89–0·95)	<0·001	0·92 (0·89–0·95)	<0·001	RLD	4·46 (3·88–5·11)	<0·001	4·56 (3·96–5·24)	<0·001

Model 1 is a base multivariable model of FEV1Q with covariates for age, sex, height, and smoking status (Former, Current, Ever, Never), plus smoking pack-years, and income to poverty ratio in NHANES and referral diagnosis for lung transplant in UNOS. Model 2 is Model 1 + an isolated race term and an interaction term between race and FEV1Q to test for both confounding and effect modification by race. Former, Ever, Never refer to smoking status. OLD: obstructive lung disease, PVD: Pulmonary Vascular Disease, CF: cystic fibrosis, RLD: restrictive lung disease.

Discussion

Amidst heightened interest in interpretive strategies for lung function that offer accurate and equitable care across races, we found FEV1Q outperforms other expressions of FEV1 in predicting mortality and is race-neutral in derivation and application. In both healthy individuals and those with advanced lung diseases FEV1Q predicted mortality and had better discriminative performance compared to currently recommended interpretation strategies. Further, across Black and White individuals there was no difference in the lower boundary of lung function approximating a survival limit, nor any difference in the performance of FEV1Q in predicting mortality. The most recent international standards and multiple recent statements from international societies recommend replacing race-specific interpretation strategies with race-neutral approaches and highlight the need for more research^1,7,36. FEV1Q has been proposed as a potential strategy that requires further investigation. These results identify FEV1Q as a race-neutral method that better predicts survival than historically used FEV1 metrics.

FEV1Q represents a conceptual shift, defining risk of mortality by comparing a patient’s FEV1 to a lower boundary approximating a survival limit rather than to a theoretical ‘normal’ value. In other words, rather than using percent of “expected” lung function, FEV1Q focuses on how much lung function remains based on a measured ‘floor’. To do so requires no reference equations, but simple arithmetic applied to an absolute FEV1 measurement, making it much easier to understand and implement. Application of FEV1Q in instances such as lung transplant evaluation, where the intention is to identify the degree of lung function remaining, is much more intuitive. Our results confirm that such a lower boundary exists and is stable across disease states, populations, and race.

The original work on FEV1Q demonstrated that the first percentile value of FEV1 settled at approximately 0·5L for males and 0·4L for females⁸. Our results confirm stable sex-specific first percentile values across all age groups of those referred for lung transplant. Values were slightly lower than those previously observed, likely reflecting the disease severity in advanced lung disease patients referred for transplant. We have now shown that these first percentile values do not differ across Black or White individuals, supporting an argument that although there may be differences in population averages of lung function by race, the lower boundary values are no different across racial groups. The previously described differences in lung function by race become evident with higher lung function, which may represent differences in the lived experience and risk factors incurred during the life course among other possible explanations. However, the association between the remaining lung function, relative to the ‘floor’, remains constant between racial groups. Because race is not necessary to define FEV1Q, it is a method for interpretation of lung function that does not propagate the concerns regarding racial inequity that have been identified in current spirometry interpretation.

Comparisons of interpretive strategies for FEV1 have been previously examined and have consistently demonstrated that absolute measures of FEV1, such as FEV1, FEV1/height², and FEV1Q, outperform methods that rely upon reference equations^12,13,37. More specific to FEV1Q, studies conducted primarily in the United Kingdom and Scandinavia have demonstrated that FEV1Q is a strong predictor of mortality across the general population and in COPD specifically^8–10,12. Our study extends these findings, examining White and Black individuals from two large cohorts, one that is representative of the general United States population, and one with disease severity reflecting end-stage lung disease, not restricted to only those with COPD. Our results address different contexts for lung function interpretation including population health and allocation of a scarce resource in lung transplant. Further we demonstrate that FEV1Q is predictive of both all-cause and respiratory-specific mortality. Together, we believe these findings provide a strong argument that FEV1Q performs better in discriminating mortality than other FEV1 interpretive strategies.

We also demonstrate a linear association between FEV1Q and mortality that is not present when using z-scores or % predicted FEV1, and that a given value of FEV1Q is compatible with a wide range of z-score values. This difference offers a putative explanation for the improved performance in discriminating mortality, namely that clinically relevant information is lost when applying reference equations to express how far FEV1 is from a predicted value compared to how far it is from a ‘bottom line’. One prior study examined this difference in associations among a smaller cohort of COPD individuals and noted a similar linear association⁹. In conjunction with the C-statistic results, these findings indicate that methods that rely on absolute values, FEV1Q in particular, more accurately define associations with survival.

While the lower boundary for FEV1 is the same irrespective of race, the median population values are lower in Black people than White people, which might suggest that survival in the Black population would be shorter than in White people based on lung function. However, rate of longitudinal lung function decline has been found to be lower in Black individuals^38–40. This suggests that there should be no difference in survival related to lung function between Black and White people. This is borne out in a prior survival study of race-neutral interpretation methods of FEV1¹⁴, and in our own findings which demonstrated no racial differences in the association between FEV1Q and survival.

The strengths of this study include that it was intentionally designed to examine healthy individuals alongside individuals with advanced lung disease, without limiting analyses to obstructive or restrictive lung disease specifically. The cohorts are large, well-characterised, and are geographically diverse as well as diverse with respect to demographics. Survival was well-defined in both cohorts allowing for accurate assessment of survival on this scale. There are also a few limitations to this study. First, all analyses were conducted using cohorts based in the United States, limiting generalizability to global populations. Second, the race-related analyses were limited to White and Black individuals as a result of underrepresentation of other minority racial groups, which is a common difficulty in many clinical research studies. This limits generalizability of these findings beyond White and Black individuals and highlights the need for more diverse representation in research studies of spirometry. Within the UNOS cohort specifically, biases related to individuals who were listed for lung transplant (e.g. those with adequate social and financial support) may also limit generalizability, and further supports the need for diverse representation in research studies of advanced lung disease. Additionally, the age distributions in both cohorts do not include children and individuals over 80 years are not well-represented. Prior work in COPD has included elderly populations and found that FEV1Q remains a significant predictor of mortality¹¹, but broader more diverse analyses in this demographic group are still required. Further, with growing recognition of the early origins of lung disease and the importance of attaining maximal peak lung function in future development of disease, similar analyses to define an equitable approach to interpretation of lung function in paediatric populations are warranted. The present analyses relied upon a single measurement of lung function and do not offer evidence with respect to longitudinal change in FEV1Q. Finally, survival was the only outcome examined in this study; acknowledging that lung function is used in various clinical settings for assessment, decision-making, and prognostication, future studies examining other outcomes of morbidity including functional status are needed.

Our results demonstrate that FEV1Q is race-agnostic and a good predictor of mortality, but there remain some considerations regarding the clinical utility of FEV1Q and further knowledge gaps that should be acknowledged. First, height is a well-established predictor of FEV1 but the associations between height and mortality, outside of those mediated by lung function are unclear. Our results demonstrate a stable lower boundary of FEV1 that approximates a survival limit regardless of height implying taller individuals live longer, and that the association between lung function and mortality was unchanged after inclusion of height in the model. However, that model also demonstrates that increased height, independent of lung function, is associated with poorer survival. Thus, the complex associations between height, lung function, and mortality remain an arena that warrants further study. Second, it should be noted that the lower boundary values of 0.4L and 0.5L for females and males respectively, do not represent a true lower limit; individuals can survive with FEV1 values lower than those values, as is evident in our own results from the UNOS cohort. Because these values are derived epidemiologically, they represent the lung function that most individuals would be able to sustain rather than a pure physiologic limitation. It is likely that the pathways by which individuals reach an FEV1 of 0.5L, contribute to their mortality more than pure ventilatory limitation. With these caveats in mind, our results demonstrate an opportunity for FEV1Q to overcome some of the existing limitations of reference equation-based interpretation strategies.

A good interpretive strategy of lung function testing needs to be accessible to providers and patients, simple to implement, support diverse patient populations and be accurate in clinically important outcomes such as survival. Our findings demonstrate that FEV1Q meets all of these criteria across two large and diverse cohorts with healthy and advanced lung disease individuals. FEV1Q is a strong prognostic indicator that does not require race for interpretation and should be considered for implementation in interpretation of pulmonary function testing.